AI-POWERED CAREER GUIDANCE AND RESUME

INTELLIGENCE SYSTEM

Project Report
Submitted to the APJ Abdul Kalam Technological University In partial of
requirements for the award of degree Bachelor of Technology
Bachelor of Technology
in
Computer Science and Engineering
by

DHRUVA T(LBT22CS044)

GOWRIPARVATHY S(LBT22CS053)

GRACE ELIZABETH JOSE(LBT22CS054)

ISHA NANDANA C S(LBT22CS058)

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

LBS INSTITUTE OF TECHNOLOGY FOR WOMEN
THIRUVANANTHAPURAM, KERALA
OCTOBER 2025
 DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING
LBS INSTITUTE OF TECHNOLOGY FOR WOMEN
THIRUVANANTHAPURAM, KERALA

CERTIFICATE
This is to certify that the report entitled ‗AI POWERED CAREER GUIDANCE AND
RESUME INTELLIGENCE SYSTEM‘ submitted by DHRUVA T (LBT22CS044),
GOWRIPARVATHY S (LBT22CS053), GRACE ELIZABETH JOSE (LBT22CS054),
ISHA NANDANA C S (LBT22CS058) to the APJ Abdul Kalam Technological University in
partial fulfillment of the requirements for the award of the Degree of Bachelor of Technology
in Information Technology is a bonafide record of the project work carried out by her under
my guidance and supervision. This report in any form has not been submitted to any other
university or institute for any purpose.

Prof. Janisha A Dr. Smitha E S Dr. Sumithra M.D

[Project Guide] [Project Coordinator] [Head of the Department]
Assistant Professor Professor Professor
Department of CSE Department of CSE Department of CSE
LBSITW, TVM LBSITW, TVM LBSITW, TVM
 DECLARATION

We undersigned hereby declare that the project report Ai-powered career guidance and
resume intelligence system submitted for partial fulfilment of the requirements for the award
of degree of Bachelor of Technology of the APJ Abdul Kalam Technological University,
Kerala is a bonafide work done by us under the supervision of Prof. Janisha A, Assistant
Professor, Department of Computer Science and Engineering, LBS Institute of Technology
for Women, Poojapura. This submission represents our ideas in our own words and where
ideas or words of others have been included, we have adequately and accurately cited and
referenced the original sources. We also declare that we have adhered to ethics of academic
honesty and integrity and have not misinterpreted or fabricated any data or idea or fact or
source in our submission. We understand that any violation of the above will be a cause for
disciplinary action by the institute and/or the University and can also evoke penal action from
the sources which have thus not been properly cited or from whom proper permission has not
been obtained. This report has not previously formed the basis for the award of any degree,
diploma, or similar title of any other University.

Place: Thiruvananthapuram DHRUVA T

Date: October 2025 GOWRIPARVATHY S
GRACE ELIZABETH JOSE
ISHA NANDANA CS
 ACKNOWLEDGEMENT

We would like to express our sincere gratitude to all the people who have guided and assisted
us through the course of this work. First and foremost, we wish to express our deep and
sincere gratitude to our Principal Dr. Smithamol M B for providing us with all the facilities
and infrastructure for the completion of this work. We record our sincere gratitude and thanks
to Dr. Sumithra M.D, Head of the Department, Department of Computer Science and
Engineering, for providing us with all the facilities for the completion of this work. We would
like to express our sincere gratitude to our project coordinator Dr. Smitha E.S, Professor,
Department of Computer Science and Engineering, for their valuable assistance provided
during the course of the project. We would also like to express our sincere gratitude to our
project guide Prof. Janisha A, Assistant Professor, Department of Computer Science and
Engineering, for providing us constant guidance, support and immense encouragement for the
successful completion of this project. We would also like to thank all the faculty members in
the department for the valuable support and encouragement they rendered. Last but not the
least, we thank all our friends and well-wishers for their kind cooperation, immense support
and useful suggestions and also for providing their truthful and illuminating views on a
number of issues related to the project.

Place: Thiruvananthapuram DHRUVA T

Date: October 2025 GOWRIPARVATHY S
GRACE ELIZABETH JOSE
ISHA NANDANA CS
 TABLE OF CONTENTS

SL NO TITLE PAGE NO

ABSTRACT i
LIST OF FIGURES ii
LIST OF ABBREVATIONS iii

1 INTRODUCTION 1-6

1.1 BACKGROUND AND MOTIVATION 1

1.2 PROBLEM DEFINITION 2

1.3 NEED FOR THE SYSTEM 3

1.4 EXISTING SYSTEM 3
1.5 GAP ANALYSIS 5
1.6 OBJECTIVES 8

2 LITERATURE REVIEW 9-16

3 SYSTEM DESIGN AND ARCHITECTURE 17-25

3.1 S Y S T E M OVERVIEW 17

3.2 SYSTEM ARCHITECTURE 18

3.3 MODULAR DESIGN OVERVIEW 20

3.4 DATA FLOW IN THE SYSTEM 22

3.5 DATASETS USED 24

3.6 ALGORITHMS AND TECHNIQUES 25

4 CONCLUSION 26-29

4.1 SUMMARY 26
4.2 FUTURE SCOPE 27

5 REFERENCES 29
 ABSTRACT

In an increasingly dynamic and competitive job market, career planning has become a
complex process that demands continuous learning, skill adaptation, and data driven decision
making. Traditional career guidance systems often provide generic suggestions, failing to
address individual skill profiles or real time industry needs. This project introduces an
intelligent Career Guidance System that leverages artificial intelligence to deliver
personalized career recommendations, automated resume analysis, and market aligned
learning suggestions helping users make informed and future ready career decisions.

The system begins by parsing user resumes to extract essential details such as education,
experience, and technical skills using advanced Natural Language Processing techniques. A
Skill Analysis and Gap Detection Engine then compares the extracted data with industry
benchmarks and evolving job role requirements to identify missing or underdeveloped
competencies. Based on this analysis, the system provides personalized course and
certification recommendations aimed at bridging identified skill gaps.

To support long term employability, a Career Roadmap Generator visualizes a user‘s optimal
career progression path, guiding them from their current state to desired professional roles.
Additionally, a Stability Prediction Module evaluates the sustainability of various roles by
analyzing market dynamics and emerging trends, offering valuable insights into job security.
The system is developed with a Flask based backend for AI model integration and a web
based frontend using HTML, CSS, and JavaScript for an interactive user experience.

By integrating intelligent resume parsing, skill benchmarking, dynamic course

recommendations, and predictive career analytics, the system serves as a comprehensive
Career GPS. It empowers students and professionals to understand their career positioning,
enhance their employability, and adapt proactively to the evolving demands of the global
workforce.

i
 LIST OF FIGURES

Fig No Title Page No

1.1 GAP ANALYSIS 10

3.1 PROPOSED SYSTEM ARCHITECTURE 18
3.2 USER INTERACTION WORKFLOW 19
4.1 MODEL COMPONENTS & TECHNIQUES 20

ii
 LIST OF ABBREVIATIONS

ABBREVIATION FULL FORM

AI Artificial Intelligence
API Application Programming Interface
BERT Bidirectional Encoder Representations from
Transformers
DL Deep Learning
EDA Exploratory Data Analysis
FNN Feedforward Neural Network
GPT Generative Pre-trained Transformer
LLM Large Language Model
ML Machine Learning
NER Named Entity Recognition
NLP Natural Language Processing
O*NET Occupational Information Network (Job Data
Source)
R² Coefficient of Determination
RAG Retrieval-Augmented Generation
RF Random Forest
RFE Recursive Feature Elimination
SMOTE Synthetic Minority Oversampling Technique
SQL Structured Query Language
SVD Singular Value Decomposition
UI User Interface
UX User Experience

iii
 CHAPTER 1
INTRODUCTION
In the era of rapidly evolving technologies and dynamic job markets, career planning has
become increasingly data-driven, personalized, and complex. With industries continuously
redefining skill requirements and the rise of interdisciplinary roles, students and professionals
often face significant challenges in identifying career paths aligned with their capabilities and
aspirations. The traditional approach of career counseling, which largely depends on static
aptitude assessments and human judgment, fails to keep pace with the constant evolution of
job landscapes and skill demands.

To address these limitations, artificial intelligence (AI) has emerged as a transformative

enabler in career guidance and decision support systems. By leveraging machine learning,
natural language processing (NLP), and knowledge retrieval techniques, it becomes possible
to analyze an individual's competencies, identify gaps, and generate intelligent, personalized
recommendations that evolve with the market. The AI-Based Career Guidance System is
designed to achieve this goal by providing a fully agentic, automated platform that evaluates
resumes, detects skill deficiencies, and recommends relevant career paths and upskilling
courses.

1.1 BACKGROUND AND MOTIVATION

In modern education and employment systems, there exists a critical disconnect between
academic learning outcomes and industry expectations. Students graduate with degrees but
often lack the practical skills demanded by recruiters, leading to employability gaps.
Employers, on the other hand, seek candidates with a precise mix of technical and soft skills,
often dynamically changing based on technological trends.

Simultaneously, the digital transformation of recruitment has resulted in massive data

generation — from job postings, resumes, and learning platforms — that can be mined for
insights. However, without intelligent systems to interpret this information, individuals are
unable to make well-informed decisions about skill development or career transitions.

1
The Career Guidance System aims to fill this gap by integrating data-driven analysis,
knowledge retrieval, and reasoning to map users‘ current skills against target job profiles. It

not only identifies missing skills but also provides actionable suggestions through course
recommendations and upskilling pathways, ultimately guiding individuals toward
employability and long-term career stability.

This approach leverages the synergy of Resume Parsing, Skill Extraction, Retrieval-
Augmented Generation (RAG) for context-aware comparison, and Feedforward Neural
Networks (FNN) for intelligent course mapping. By adopting these technologies, the system
moves beyond static recommendation systems toward an adaptive, learning-based framework
capable of evolving with user data and market dynamics.

1.2 PROBLEM DEFINITION

Despite the availability of numerous job portals and e-learning platforms, most existing
systems operate in isolation. Job recommendation systems are primarily keyword-based,
while course platforms recommend generic content unrelated to an individual‘s professional
trajectory. Furthermore, there is a lack of intelligent systems that can holistically analyze a
person‘s resume, compare it against role-specific requirements, and suggest actionable
improvement plans.

The absence of integration between resume understanding, skill benchmarking, and career
path optimization creates inefficiency in talent development and employability enhancement.
Manual career counseling is time-consuming and subjective, whereas automated systems
without AI reasoning lack contextual accuracy.

Thus, the problem addressed by this project is to design and implement a comprehensive, AI-
driven career guidance platform that intelligently interprets user profiles, identifies skill gaps,
and generates targeted career recommendations by integrating Retrieval-Augmented
Generation (RAG), Feedforward Neural Networks (FNN), and Large Language Model
(LLM) reasoning capabilities.

2
1.3 NEED FOR THE SYSTEM

The need for an intelligent career guidance system arises from multiple challenges observed
in both educational and employment sectors:

1. Skill-Job Mismatch: Graduates often lack awareness of the skills required for specific
roles, leading to poor alignment with market needs.
2. Static Counseling Approaches: Manual career guidance is subjective and cannot
dynamically adapt to changes in job trends or skill relevance.
3. Overwhelming Data Availability: With millions of job postings and courses online,
users struggle to identify what truly suits their goals.
4. Lack of Personalized Recommendations: Existing platforms lack the ability to generate
personalized, skill-based insights derived from resume data.
5. Limited Market Awareness: Without understanding evolving trends, users may upskill
in irrelevant domains, reducing employability.

The Career Guidance System directly addresses these issues through an integrated and
intelligent framework. It enables dynamic, real-time benchmarking of user profiles against
industry expectations, offering data-driven and adaptive guidance that continuously evolves
with market changes.

1.4 EXISTING SYSTEM

The existing career advisory landscape is primarily dominated by fragmented systems that
operate independently across different functions such as job search, skill learning, and resume
evaluation. While several online platforms provide partial support in these areas, none offer
an end to end intelligent solution that analyzes individual profiles, identifies skill
deficiencies, and provides personalized recommendations to enhance employability.

Most job seeking and learning systems today fall into three broad categories: job
recommendation platforms, resume screening tools, and online learning or course
recommendation portals. Each category plays a valuable role, yet their lack of integration
results in an inefficient and incomplete career guidance ecosystem.

3
 Job Recommendation Platforms
 Existing job platforms like LinkedIn, Naukri, and Indeed rely on keyword-based
matching that overlooks the context of a candidate‘s skills, experience, and goals,
leading to generic and shallow recommendations without identifying skill gaps or
providing meaningful career insights.
 Resume Screening and Analysis Tools
 Automated resume screening tools like ATS, Rezi, or Resume Worded focus on
formatting and keyword density but lack contextual understanding of a candidate‘s
true capabilities, expertise level, and learning potential, offering limited value for
personalized career guidance or upskilling recommendations.
 Online Learning and Course Recommendation Platforms
 E-learning platforms like Coursera, Udemy, and edX offer extensive course libraries
but rely on popularity-based suggestions rather than personalized, skill-gap–driven
recommendations, often leading learners to pursue courses misaligned with their
career goals and employability needs.

 Institutional Career Counseling

 At the academic level, career counseling remains manual and subjective, relying on
limited faculty insights rather than data-driven analysis, lacking scalability, real-time
labor market alignment, and evidence-based mechanisms to guide students toward
stable and employable career paths.

Limitations of Existing Systems

1. Lack of Integration:
 Current systems function in isolation. Job search engines, resume analyzers, and
learning platforms operate independently, leaving users to manually connect insights
across them.
2. Absence of Intelligent Reasoning:
 The lack of AI driven contextual understanding prevents meaningful interpretation of
a candidate‘s potential or career trajectory.
3. Static Skill Matching:
 Keyword based algorithms cannot adapt to evolving market skill requirements or
personalized goals.

4
4. No Skill Gap Benchmarking:
 Existing systems do not evaluate the difference between a user‘s current skill set and
the ideal profile for a target role.
5. Limited Personalization:
 Recommendations are often generic, ignoring user preferences, learning patterns, and
professional aspirations.
6. No Predictive Insights:
 There is no mechanism to forecast career stability, emerging job domains, or future
skill relevance.

1.5 GAP ANALYSIS

Existing career guidance approaches, across job portals, resume analyzers, e-learning
platforms, and academic counseling,lack personalization, contextual understanding, and data-
driven insights, offering generic and disconnected recommendations. To overcome these
limitations, the proposed AI-Based Career Guidance System integrates NLP, RAG, and
machine learning within a unified framework to provide intelligent, adaptive, and evidence-
based career recommendations tailored to each student‘s profile and evolving industry trends.

 Lack of Contextual Understanding in Resume Evaluation

Most resume analyzers rely on surface-level keyword extraction without understanding

context or skill depth, treating all mentions of terms like Python or machine learning as equal.
This lack of semantic reasoning leads to false equivalence and generic feedback, preventing
accurate assessment of a candidate‘s proficiency or readiness for specific roles.

 Absence of Dynamic Skill Gap Benchmarking

Existing job platforms and resume analyzers cannot quantify the difference between a user‘s
current competencies and the skill requirements of a desired job role. Without such
benchmarking, candidates remain unaware of which specific skills to acquire or how far they
are from achieving employability in their target domain.

5
 Lack of Personalized Course Recommendations

Most e-learning platforms recommend courses based on popularity, previous user activity, or
static category mapping rather than individual learning needs. Such systems fail to align
learning resources with the user‘s actual deficiencies or career trajectory.

 Fragmented Systems with No Unified Architecture

Existing systems function independently—resume analyzers, job portals, and e-learning sites
do not share data or insights. As a result, users receive disconnected feedback that lacks
actionable coherence. This fragmentation leads to redundant efforts and prevents users from
forming a continuous improvement loop between skill assessment and upskilling.

 Static and Non-Adaptive Recommendations

Traditional systems lack adaptability to evolving industry trends. Once trained, their rule-
based or static algorithms fail to update automatically as new job roles and technologies
emerge.
This results in outdated suggestions that no longer reflect real-world job market needs.

 No Predictive Insights on Career Stability

While current systems provide skill or course recommendations, none assess the long-term
stability or growth potential of a career path. Users are left without guidance on whether their
chosen roles are future-proof or at risk due to automation or market shifts.

 Limited Explainability and User Awareness

Another major gap lies in the lack of transparency in AI-driven recommendations. Existing
systems provide results without explaining why a particular job or course is recommended,
reducing user trust and interpretability.

6
Summary of Identified Gaps

Category Gap in Existing System Proposed Solution

Keyword-based, non- Contextual NLP + LLM-driven

Resume Analysis
contextual parsing understanding

No quantitative skill gap

Skill Benchmarking RAG-based skill gap matrix
identification

Course
Generic or popularity-based FNN + personalized mapping
Recommendation

System Integration Disconnected platforms Unified Flask-based AI architecture

Dynamic hybrid RAG + GPT-4o

Adaptability Static, rule-based models
reasoning

No stability or future scope

Predictive Insights AI-based career stability prediction
analysis

Natural-language explanations and

Explainability Black-box outputs
transparency

FIG 1.1:GAP ANALYSIS

In conclusion, the gap analysis highlights that current systems fall short of delivering a
comprehensive, adaptive, and interpretable AI-driven career guidance ecosystem. The AI-
Based Career Guidance System bridges these gaps through an integrated approach that
combines resume interpretation, contextual reasoning, intelligent recommendation, and
predictive career analytics, providing users with an informed, future-oriented pathway toward
employability and growth.

1.6 OBJECTIVES

The AI-Powered Career Guidance and Resume Intelligence System aims to develop an
intelligent, data-driven solution that assists individuals in identifying their current career
7
standing, skill deficiencies, and future opportunities through AI-based analysis and predictive
modeling. The system integrates advanced technologies such as Natural Language Processing
(NLP), Retrieval-Augmented Generation (RAG), and Machine Learning (ML) to provide
accurate, personalized, and adaptive recommendations for career development.

The project has been designed around five major objectives, each addressing a distinct
challenge in the existing career guidance ecosystem.

 Develop an Intelligent Resume Parsing Engine

To accurately extract structured data such as skills, education, and experience from
unstructured resume formats using NLP and GPT-based parsing.

 Perform Automated Skill Gap Analysis

To benchmark current user skills against the requirements of target job roles using Retrieval-
Augmented Generation (RAG) and GPT-4o reasoning.

 Generate Personalized Learning Recommendations

To recommend relevant courses and certifications using an AI model (Feedforward Neural

Network) integrated with open learning platforms such as MIT OCW, NPTEL, and
freeCodeCamp.

 Design Career Roadmaps

To construct dynamic, goal-oriented career progression paths that visually guide users toward
achieving their desired roles.

 Estimate Job Role Stability

To predict job role demand and layoff probabilities through analysis of external datasets like
[Link] and LinkedIn workforce analytics.

8
 CHAPTER 2
LITERATURE REVIEW
Paper 1: “Research on Data Driven Student Personalized Learning Path Recommendation
System — Y Sihong” (IEEE ICIPCA 2025)

Y Sihong introduced a personalized learning recommendation system that integrates data

driven modeling with adaptive learning optimization to generate individualized learning
paths. The study employed an Adaptive Graph Convolutional Network AGCN combined with
Reinforcement Learning RL to model the relationships between learning units, skills, and
student progress. The AGCN was responsible for constructing a graph based knowledge
structure, capturing interdependencies among various topics and skills, while RL dynamically
adjusted the learning sequence based on the learners feedback, historical performance, and
goal completion metrics.

This hybrid approach enabled the system to learn from past user behavior and continuously
improve the recommendation strategy. The model achieved an average improvement of 15 to
18 percent in learning path accuracy and completion rates compared to static
recommendation systems. The experiments were conducted using an educational dataset with
more than 5000 student profiles, providing strong empirical validation.

Despite these promising results, the research highlighted certain challenges. The system
required a large amount of labeled training data and struggled to generalize across
heterogeneous learners due to demographic and academic diversity. Moreover, the models
high computational demand made it difficult to deploy in real time low resource
environments such as university career guidance cells.

The findings of this paper directly influence the Personalized Learning Recommendation
Module of the proposed system. Similar to Sihongs adaptive mechanism, your projects
Feedforward Neural Network FNN and RAG based retrieval logic dynamically adapt
recommendations according to user progress and skill gaps. The principle of feedback based
adaptability from this research serves as the conceptual backbone for the intelligent
recommendation layer in your project, ensuring the system evolves continuously with the
users growth and industry requirements.

9
Paper 2 : AI Driven Intelligent Career Mapping and Skill Based Job Role Predicting
Career Guidance System — K Selvaraj et al (IEEE ICAISS 2025)

K Selvaraj and colleagues proposed a comprehensive AI framework for career role prediction
and skill based job recommendation focusing on the automation of career counseling
processes. Their model employed a two stage prediction pipeline a Random Forest classifier
for coarse grained domain prediction and a Light Gradient Boosting Machine LightGBM for
fine grained job role classification. The Random Forest model was trained on 2500 labeled
data points containing attributes such as educational background, technical skills,
certifications, and interests. The LightGBM model then refined predictions by identifying the
most likely job roles within the predicted domain, yielding a top 3 recommendation list.

The system was tested on a student dataset collected through surveys, achieving 90.5 percent
accuracy in predicting job categories and 87 percent accuracy in recommending roles. It also
included a task completion tracker that encouraged users to engage in skill development
activities based on model suggestions. However, the system relied heavily on manually
collected and static data, making it less responsive to real world job market fluctuations. The
absence of semantic understanding and proficiency level differentiation among skills reduced
recommendation depth.

This study forms the conceptual basis of the Career Guidance Systems career mapping
component, where the static rule based model is replaced with a context aware reasoning
mechanism. By employing Retrieval Augmented Generation RAG and GPT 4o, the proposed
project enhances adaptability, retrieves live job role data from curated repositories, and
generates semantically grounded recommendations. Unlike Selvarajs approach, your system
interprets user resumes through NLP and continuously updates job mappings to reflect
dynamic industry expectations.

Paper 3: “A Machine Learning Based Employee Layoff Prediction System — C Vasuki et

al” (IEEE AIMLA 2025)

This research introduces a sophisticated layoff prediction framework designed to forecast

employee retrenchments using multi dimensional data sources. C Vasuki and team developed
an Advanced Random Forest Algorithm RFA, enhanced with Recursive Feature Elimination
RFE and Synthetic Minority Oversampling Technique SMOTE to counter data imbalance
issues prevalent in HR datasets. The model integrated a wide range of features, including

10
employee performance indicators, project completion rates, tenure duration, company
profitability, and macroeconomic metrics such as GDP variation and industry growth rates.

The study reported an outstanding accuracy rate of 99.5 percent with an F1 score of 0.97,
demonstrating the models robustness and reliability. It also incorporated an interactive Flask
based dashboard, which visualized risk probabilities and contributing factors, thus improving
interpretability for HR decision makers. Despite its strong technical design, the model was
restricted to organizational use focusing on employer analytics rather than individual
employee risk awareness or prevention strategies.

The insights from this work directly influence your Career Stability Prediction Module,
where the same concept is applied at an individual level. In your system, predictive analytics
are used not to evaluate layoffs from a companys perspective but to assess the stability and
sustainability of job roles in the broader labor market. By combining macroeconomic trend
data and skill redundancy indicators, your system offers users personalized stability insights,
empowering them to make data driven career choices with awareness of future employment
risks.

Paper 4:” Optimizing Resume Parsing Processes by Leveraging Large Language Models,
V Manish et al”( IEEE TENSYMP 2024)

In this paper, V Manish et al explored the use of Large Language Models LLMs to enhance
the efficiency and accuracy of resume parsing processes. The study aimed to overcome the
limitations of rule based systems that fail to handle unstructured and inconsistent resume
formats. The researchers employed prompt engineering and contextual schema validation,
using an LLM to identify and extract entities such as education, technical skills, work
experience, and achievements. The extracted data was validated against predefined schema
constraints to ensure uniformity and minimize hallucination errors.

Experimental comparisons revealed that the LLM based parser achieved 25 percent higher
field extraction accuracy than conventional methods like spaCy and regex parsing,
particularly for resumes with inconsistent formatting or graphical elements. However, the
study noted issues of computational overhead, inconsistent response patterns, and high token
processing costs, which posed scalability challenges for deployment at scale.

Your Intelligent Resume Parsing Engine draws strong influence from this work but improves
upon it through a hybrid NLP LLM architecture. Instead of relying exclusively on an LLM,
11
the system integrates spaCy for initial rule based extraction and GPT 4o for contextual
understanding. This significantly reduces latency while retaining the interpretive benefits of
large models. Furthermore, the hybrid model adds an error checking layer to detect
hallucinations and applies semantic normalization for consistent skill representation, making
it more suitable for real world career analytics applications.

Paper 5: “Resume Ranker AI Based Skill Analysis and Skill Matching System ,N Gangoda
et al” ( IEEE ICDS 2024)

The Resume Ranker proposed by N Gangoda and colleagues focuses on automated skill
analysis and resume job matching using NLP and vector based similarity techniques. The
methodology involved extracting keywords and phrases using KeyBERT and Word2Vec,
representing both resumes and job descriptions in vector space, and calculating cosine
similarity to determine the degree of alignment. The system also provided basic course
suggestions based on missing skills and ranked resumes accordingly.

Although effective in producing ranked lists and simple feedback, the model lacked
contextual understanding. It could not distinguish between similar skills in different domains
for example data analysis in marketing vs engineering. Furthermore, the system failed to
incorporate market dynamics, meaning the recommendations quickly became outdated as job
trends evolved.

This work serves as a technical foundation for your Skill Gap Analyzer, which extends the
concept from keyword level comparison to semantic reasoning using Retrieval Augmented
Generation RAG. The proposed module does not just match skills; it interprets the context,
measures proficiency levels, and visualizes deficits in a Skill Gap Matrix. By grounding
recommendations in real time data retrieved from curated sources, your system achieves a
level of accuracy and contextual adaptability far beyond the scope of the original Resume
Ranker.

Paper 6: AI Powered Model for Intelligent Resume Recommendation and Feedback ,A.
Mishra et al. (IEEE ICCCNT, 2024)

A. Mishra and collaborators developed a comprehensive AI powered resume

recommendation and feedback system that leverages Bidirectional Encoder Representations
from Transformers (BERT) to establish semantic relationships between resumes and job
descriptions. The core idea revolves around encoding both resumes and job postings into
12
dense contextualized vectors within a shared embedding space. Once represented, cosine
similarity is used to compute the semantic closeness between the candidate profile and job
requirements. Unlike conventional keyword based approaches, this method captures deeper
contextual nuances, enabling the system to identify implicit relationships between skills and
job expectations.

The model was evaluated using the Kaggle Resume Dataset and achieved a substantial
improvement in relevance accuracy, registering up to 30 percent higher alignment scores than
traditional TF IDF and Bag of Words approaches. Additionally, the model provided candidate
feedback by identifying missing or weakly represented skills. However, the paper also
acknowledged that the model lacked a proficiency metric for each extracted skill, meaning it
treated all skills as equally weighted. Furthermore, the absence of a temporal understanding
of skill usage limited its ability to assess the recency or relevance of experience.

The proposed Career Guidance System expands upon this study by introducing a quantitative
Skill Gap Analyzer that not only identifies missing skills but also benchmarks them based on
industry relevance and individual proficiency. The system integrates RAG based retrieval and
GPT 4o semantic reasoning, ensuring that each recommendation is contextualized within real
time market expectations. This enhancement transforms the simple feedback loop of Mishra‘s
model into a multi layered analytical framework capable of adaptive reasoning and actionable
skill progression mapping.

Paper 7: Analyzing Trends, Skills Demand, and Salary Prediction in the AI and ML Job
Market — D. Ather et al. (IEEE IIPEM, 2024)

In this study, D. Ather et al. explored the global AI and Machine Learning job market,
focusing on demand analysis, skill evolution, and salary forecasting through machine
learning. The researchers used a dataset compiled from online recruitment platforms such as
Indeed and Glassdoor, covering over 75,000 job postings between 2022 and 2024. The data
included job titles, salary ranges, skill requirements, and geographic locations.

To analyze this data, the team applied XGBoost Regression, Random Forest, and Support
Vector Machines (SVMs) to predict average salary values based on skill combinations and
experience levels. They achieved an R squared value of 0.82, demonstrating strong predictive

13
reliability. Moreover, keyword frequency analysis revealed that emerging technologies such
as Cloud Computing, Data Engineering, and AI Deployment Pipelines have seen exponential
growth in demand. The study also highlighted a shift toward hybrid and remote job
opportunities, reflecting post pandemic workforce evolution.

Despite its strong findings, the model‘s geographical limitation, being based largely on U.S.
data, restricted its global applicability. Furthermore, the skill extraction pipeline was keyword
centric, lacking semantic representation and contextual reasoning. These limitations make it
difficult to generalize or update predictions as job descriptions evolve.

The proposed Career Stability Prediction Module in your system draws heavily from this
research by integrating multi source market data aggregation and retrieval enhanced
forecasting. By incorporating dynamic datasets from open repositories and applying AI
driven stability scoring, your model evaluates not just salary predictions but also the long
term sustainability of specific roles. This enables users to understand both short term
employability and long term career security, bridging a critical gap in the literature.

Paper 8: Research on Personalized Learning Recommendation Model Based on User

Collaborative Filtering Algorithm ,Y. Cheng (IEEE RAIIC, 2024)

Y. Cheng proposed a hybrid learning recommendation system that integrates user based
collaborative filtering with content based filtering to improve personalization in e learning
environments. The model captures user similarities based on historical learning behaviors,
preferences, and performance patterns. By combining this with content similarity derived
from course metadata, it provides a more balanced recommendation strategy.

The dataset for evaluation consisted of over 20,000 user interaction records sourced from a
university‘s online learning management system (LMS). The hybrid model achieved a top N
recommendation accuracy improvement of 12 percent over traditional collaborative filtering
approaches. However, Cheng‘s system faced significant challenges in handling unstructured
data, such as textual inputs from resumes or non standard learning profiles. Moreover, it
suffered from the cold start problem, where the absence of historical data for new users led to
poor recommendations.

14
This research forms a conceptual backbone for your Learning Recommendation Module,
which replaces the dependency on user history with RAG based reasoning and Feedforward
Neural Networks (FNN) that rely on extracted skill data rather than behavioral patterns. The
integration of resume derived features enables your system to operate effectively even for
first time users with no prior activity. Furthermore, your system dynamically updates
recommendations as users acquire new skills, effectively overcoming the static limitations
observed in Cheng‘s framework.

Paper 9: Machine Learning Based Ideal Job Role Fit and Career Recommendation
System, S. K. S (IEEE ICCMC, 2023)

S. K. S and his team introduced a matrix factorization based recommendation system for
identifying optimal job roles for individuals. The authors experimented with Singular Value
Decomposition (SVD), SVD++, and Non Negative Matrix Factorization (NMF) to uncover
hidden relationships between user profiles and job roles within high dimensional data. The
study was conducted using the Stack Overflow Developer Survey dataset, which contains
over 80,000 samples representing developers‘ skills, experience levels, and preferred
technologies.

Among the tested algorithms, SVD++ delivered the highest precision (0.89) and recall (0.86),
primarily due to its ability to model implicit feedback, capturing both what users explicitly
stated and what could be inferred from their interactions. The model also demonstrated the
capability to generalize across roles by identifying latent similarities in skill requirements.
However, due to data sparsity and the static nature of matrix factorization, the system
struggled with adaptability and required retraining when new job roles or technologies
emerged.

Your Career Guidance System builds upon this foundation by replacing static embeddings
with dynamic neural representations and retrieval enhanced embeddings. Unlike matrix
factorization, which relies on fixed latent vectors, your model learns continuously from live
job postings, enabling adaptive role fit analysis that evolves with market changes. By
combining retrieval based augmentation with neural contextualization, the system ensures
long term scalability and accurate, explainable recommendations across different professional
domains.

15
Paper 10: Fast and Accurate Resume Parsing Method Based on Multi Task Learning ,Y.
Liang et al. (IEEE IALP, 2023)

Y. Liang and co researchers presented a multi task learning model for resume parsing that
jointly performs topic classification and Named Entity Recognition (NER). The architecture
integrates BERT embeddings with a Bidirectional Long Short Term Memory (BiLSTM) layer
followed by a Conditional Random Field (CRF) decoder. This combination allows the model
to capture both global contextual meaning and local sequential dependencies. By sharing
parameters across the two tasks, the model achieved higher efficiency and reduced
computational redundancy.

The experiments used a multilingual dataset containing English, Chinese, and Japanese
resumes. The model achieved an impressive F1 score of 0.93 for entity recognition and a
processing time reduction of 48 percent compared to traditional sequential task models. The
results confirm the effectiveness of multi task learning for structured information extraction.
However, the paper acknowledges the limitation of requiring large annotated datasets and its
inability to normalize extracted skills across industries.

The Intelligent Resume Parsing Engine in your project takes direct inspiration from Liang‘s
research. It adopts a similar joint learning concept by combining spaCy based rule extraction
and GPT 4o semantic interpretation in a two stage pipeline. This hybrid design enhances
multilingual compatibility and schema normalization, ensuring that resume data is not only
extracted accurately but also organized for downstream processing like Skill Gap Analysis
and Career Mapping. The integration of explainable LLM outputs further increases
interpretability, making your parsing system both scalable and user transparent.

16
 CHAPTER 3
SYSTEM DESIGN AND ARCHITECTURE

The design of the AI Powered Career Guidance and Resume Intelligence System follows a
modular, layered, and data driven architecture that enables efficient integration of multiple AI
components. The goal of the design is to create an intelligent, interactive platform that
automates resume analysis, identifies skill gaps, recommends personalized learning paths,
generates career roadmaps, and predicts the stability of job roles.

The system is implemented using a Flask based backend for orchestration and
communication between modules and a web based frontend (HTML, CSS, and JavaScript)
for user interaction. Each module operates autonomously but communicates through a central
coordination layer to ensure scalability and fault isolation. The architecture emphasizes
semantic understanding, real time reasoning, and modularity, allowing for future expansion
and integration with additional data sources.

3.1 SYSTEM OVERVIEW

The Career Guidance System is designed as a five module AI pipeline where each stage
contributes a specific function toward intelligent employability analysis. The workflow
begins with resume upload and ends with personalized guidance output presented through an
intuitive dashboard.

1. Input Layer: Users upload their resumes in PDF DOCX or image format through the
frontend interface.
2. Processing Layer: The resume is preprocessed for text extraction followed by
contextual data structuring using AI models.
3. Analysis Layer: Skills are identified benchmarked and compared against target job
roles using NLP and Retrieval Augmented Generation RAG.
4. Recommendation Layer:Based on skill deficiencies the system generates adaptive
learning recommendations and career roadmaps.
5. Prediction Layer The model predicts the stability and long term relevance of the target
role using historical trend data.

17
The modular nature of the system ensures that improvements or upgrades in one module for
example resume parsing do not affect the functionality of others. The end result is an
intelligent guidance framework that learns continuously from user interaction and evolving
job market trends.

3.2 SYSTEM ARCHITECTURE

The system follows a multi layered architecture integrating several AI driven modules
through a centralized Flask based backend. It can be divided into five major layers.

1. User Interface Layer

This layer represents the user facing component of the system developed using HTML CSS
and JavaScript. It allows users to

 Upload resumes in supported formats

 View extracted profile data identified skill gaps and generated insights
 Access personalized learning recommendations career roadmaps and stability
visualizations

The frontend is designed with simplicity and interactivity in mind ensuring seamless user
engagement.

2. Application Backend Layer

The Flask backend acts as the systems central control unit. It coordinates communication
among AI modules and manages requests between the frontend and backend. Flask handles

 Routing and API communication

 Data preprocessing and task assignment to AI modules
 Integration with machine learning pipelines
 Session and user data management for consistent tracking

This modular backend structure ensures flexibility scalability and the ability to integrate
additional AI models in future iterations.

18
 FIG 3.1 PROPOSED SYSTEM ARCHITECTURE

3. AI Processing Layer

This is the core intelligence layer of the system responsible for analytical and predictive
operations. It comprises the following submodules

1. Resume Parsing Engine Extracts and structures information from uploaded resumes
using spaCy,PyPDF2 and GPT4o.
2. Skill Gap Analyzer Benchmarks the users current skills against job role datasets using
RAG based retrieval and semantic comparison.
3. Learning Recommendation Engine Utilizes a Feedforward Neural Network FNN
integrated with RAG retrieval to suggest relevant open access learning materials.
4. Career Roadmap Generator Constructs dynamic stepwise career progression paths
based on current and target skills.
5. Career Stability Predictor Applies supervised learning models Random Forest
LightGBM to forecast job role sustainability and long term viability.

19
Each module operates as an independent service within the backend ecosystem
communicating through lightweight JSON data exchange protocols.

4. Database Layer

The database maintains persistent storage for user information extracted resume data skills
job role taxonomies and recommendation history. A MySQL or SQLite database is used
structured into relational tables such as:

♦ user_profile
♦ parsed_resume
♦ skills_matrix
♦ learning_resources
♦ stability_scores

This allows efficient querying and retrieval for continuous user interaction and long term data
driven improvement.

5. Knowledge and Data Source Layer

This layer integrates external and internal datasets for contextual reasoning and retrieval. It
includes

 Curated job role datasets and skill ontologies

 Market trend datasets for stability prediction
 Free learning repositories and institutional knowledge bases
 RAG connected knowledge stores for contextual information retrieval

By combining structured and unstructured data sources this layer empowers the systems
adaptive reasoning and contextual accuracy.

3.3 MODULAR DESIGN OVERVIEW

The Career Guidance System is structured into five interdependent modules each addressing
a key stage of the analytical pipeline.

20
1. Resume Parsing Module

This module automates the extraction of structured data from resumes using PyPDF2 and
spaCy NER for basic entity recognition. For complex or ambiguous text GPT4o assists by
providing context aware understanding. It identifies education technical and soft skills
projects certifications and experiences. Parsed data is converted into a structured JSON
format to enable downstream processing.

2. Skill Gap Analyzer

The Skill Gap Analyzer serves as the analytical core of the system. It benchmarks the
extracted skills against industry expectations retrieved using Retrieval Augmented Generation
RAG. By combining contextual retrieval with LLM reasoning it identifies missing or
underdeveloped competencies and quantifies them into a Skill Gap Matrix. Each gap is
ranked based on its importance to employability in the users desired role.

3. Learning Recommendation Engine

Using the generated skill gap data the Recommendation Engine identifies relevant free and
open access learning resources. The module uses an FNN model that maps skill deficiencies
to appropriate educational materials retrieved from knowledge bases. It ensures that the
learning recommendations are both contextually appropriate and personalized to each users
career goal.

4. Career Roadmap Generator

This module constructs dynamic visualized career paths that depict the users current standing
targeted roles and required upskilling milestones. Using job role hierarchies and dependency
data retrieved via RAG it produces a progression model outlining which skills to acquire in
what sequence. This helps users visualize a structured journey toward employability and
career advancement.

5. Career Stability Predictor

The final module performs predictive analysis on job roles using machine learning models
such as XGBoost and LightGBM. It evaluates long term career sustainability by analyzing
21
job market trends automation risks and demand trajectories. The output a stability score
informs users about the future security of their chosen career paths encouraging proactive
learning and diversification.

3.4 Data Flow in the System

The data flow within the Career Guidance System follows a well defined sequence

1. The user uploads a resume

2. The backend extracts and normalizes the data
3. Skills are analyzed and compared against industry data
4. Missing skills are identified and the recommendation module generates upskilling
suggestions
5. The system builds a personalized roadmap and predicts job stability
6. All processed results are sent back to the frontend dashboard

This flow ensures that the system maintains a feedback driven loop enabling continuous
improvement and user engagement.

22
 23
FIG 3.2:USER INTERACTION FLOW
The User Interaction Flow describes the sequence of operations from the user‘s perspective
and shows how the front-end and back-end components communicate to generate insights.

As shown in Figure 3.2, the process begins with the user accessing the web-based interface
developed using HTML, CSS, and JavaScript. After logging in, the user uploads their
resume, which is processed by the Flask-based backend server. The backend triggers the AI
modules like Resume Parsing, Skill Gap Analysis, Course Recommendation, Roadmap
Generation, and Stability Prediction each contributing results that are stored in the SQLite
database. The system then generates visual dashboards that display skill gaps, learning paths,
and job stability analytics in a consolidated format.

This interaction design ensures ease of use, intuitive navigation, and minimal user
intervention, making the entire process efficient and user-friendly.

3.5 DATA SETS

The system utilizes both custom-built and open-source datasets for training and evaluation of
various modules:

 Resume Dataset: A curated collection of anonymized resumes from publicly available

repositories (e.g., Kaggle Resume Dataset) used for training the parsing and embedding
models.
 Job Role Dataset: Extracted from LinkedIn Jobs, O*NET, and Indeed, containing details
about job titles, required skills, and industry-level trends.
 Course Dataset: Compiled from MIT OCW, NPTEL, and freeCodeCamp, including
course titles, tags, prerequisites, and skill mappings.
 Labor Market Dataset: Real-time data sourced from [Link] and online economic
APIs, containing company layoff events, workforce changes, and job demand indices.
All datasets undergo preprocessing to remove redundancies, unify skill naming conventions,
and normalize metadata for consistent representation.

24
3.6 ALGORITHMS AND TECHNIQUES
The proposed system integrates a variety of algorithms from the domains of machine
learning, NLP, and deep learning, each optimized for its corresponding module:

Module Primary Technique Purpose

Resume Parsing Contextual skill and entity
NLP (spaCy) + GPT-4o + OCR
Engine extraction

Retrieval-Augmented Generation Benchmarking user skills against

Skill Gap Analyzer
(RAG) job requirements

Course
Feedforward Neural Network Ranking and recommending free
Recommendation
(FNN) + Cosine Similarity learning resources
Engine

Career Roadmap Generating adaptive, stepwise

LLM-based Conditional Reasoning
Generator growth trajectories

Job Stability Predicting job security and

Regression and Sentiment Analysis
Prediction Engine stability scores

FIG 3.3:MODEL COMPONENTS AND TECHNIQUES

Additionally, the system employs vector embeddings (BERT/Word2Vec) for skill

representation, SQLite queries for data management, and Flask REST routing for
communication between modules.

The architectural philosophy emphasizes scalability, modularity, and interpretability,

ensuring that future extensions—such as generative resume enhancement or interview
readiness analytics—can be integrated seamlessly.

25
 CHAPTER 4
CONCLUSION

4.1 SUMMARY

The project titled “AI-Powered Career Guidance and Resume Intelligence System”
presents a comprehensive approach to bridging the gap between individual capabilities and
dynamic industry demands through artificial intelligence, machine learning, and natural
language processing. The system aims to function as an intelligent career companion, capable
of parsing resumes, analyzing skills, identifying deficiencies, recommending learning paths,
visualizing career roadmaps, and predicting job stability all through a unified digital platform.

In Phase 1, the focus was placed on a thorough conceptual and architectural foundation for
the system. This included a comprehensive literature review of state-of-the-art methodologies
in skill recommendation, resume parsing, semantic analysis, and job market prediction. Each
reviewed paper provided valuable insights that guided the selection of algorithms and design
strategies, ensuring the proposed solution remains technically sound and practically feasible.

The system design phase successfully established the project‘s multi-layered architecture
encompassing five key modules: the Resume Parsing Engine, Skill Gap Analyzer, Course
Recommendation Engine, Career Roadmap Generator, and Job Stability Prediction Engine.
Each component was mapped to specific AI methodologies such as spaCy NLP, GPT-4o-
based reasoning, Feedforward Neural Networks, and Retrieval-Augmented Generation
(RAG), forming an integrated analytical pipeline. The design ensures scalability, modularity,
and interpretability, allowing for future enhancements without disrupting the system‘s
workflow.

Phase 1 also emphasized data-driven adaptability and user-centric design principles. Unlike
traditional systems that offer static recommendations, this system proposes dynamic and
evolving insights that adapt as the user progresses. By focusing on free educational resources
(like MIT OCW, NPTEL, and freeCodeCamp) and real-time labor market intelligence (from
sources such as [Link] and LinkedIn trends), the project addresses both accessibility and
industry relevance.

Through this phase, the groundwork has been firmly laid for developing a practical, AI-
enabled framework that not only evaluates where a user stands today but also predicts where

26
they can optimally grow tomorrow. The completion of Phase 1 thus establishes a solid
structural and theoretical foundation, upon which the implementation, testing, and
performance validation of the proposed system will be built in the subsequent project phase.

4.2 FUTURE SCOPE

The next phase of the project will focus on the implementation, testing, and optimization of
the proposed system. The following developments are planned to enhance its capabilities and
extend its real-world applicability:

1. System Implementation
Each designed module including Resume Parsing, Skill Gap Analysis, Course
Recommendation, and Job Stability Prediction will be implemented using Python, Flask,
and NLP frameworks such as spaCy and LangChain. The frontend will be developed
using HTML, CSS, and JavaScript, providing an intuitive and interactive dashboard for
users.
2. Integration of Machine Learning Models
The Feedforward Neural Network (FNN) for course recommendations and regression
models for stability prediction will be trained and fine-tuned using curated datasets. Real-
time performance evaluation will be conducted to ensure accuracy and response
efficiency.
3. Database and API Development
A SQLite-based data management layer will be implemented to store and retrieve parsed
resume information, user progress, and AI-generated insights. Future versions may
migrate to PostgreSQL or MongoDB for scalability and cloud integration.
4. User Feedback Loop
A continuous feedback mechanism will be integrated, allowing users to rate
recommendations and learning outcomes. This data will be used to retrain models
periodically, ensuring continuous learning and adaptive improvement.
5. Enhanced Visualization Dashboard
An interactive dashboard will be designed using Plotly and Matplotlib, providing visual
summaries of skills, progress metrics, and predicted job stability trends. The interface will
aim to make complex analytics comprehensible to non-technical users.
6. Performance Evaluation and Validation

27
 The system will undergo module-level and integrated testing, including accuracy,
precision, recall, and usability evaluations. Comparative analysis against existing
platforms (like LinkedIn Learning and Coursera Career Tools) will be conducted to assess
real-world competitiveness.
7. Security and Ethical Considerations:
Future enhancements will include data privacy safeguards, role-based authentication, and
ethical AI compliance to ensure user confidentiality and unbiased recommendations.
8. Scalability and Cloud Deployment
In the extended scope, the system can be migrated to a cloud-based environment to
handle multiple concurrent users efficiently, enabling deployment as a web service for
educational institutions and career platforms.

28
 REFERENCES
1) .Sihong, "Research on Data-Driven Student Personalized Learning Path
Recommendation System," 2025 IEEE 3rd International Conference on Image
Processing and Computer Applications (ICIPCA), Shenyang, China, 2025.
2) K. Selvaraj, V. K, M. R. P and M. A. M, "AI-Driven Intelligent Career Mapping and
Skill-based Job Role Predicting Career Guidance System," 2025 Third International
Conference on Augmented Intelligence and Sustainable Systems (ICAISS)
3) C. Vasuki, G. M S, D. M and D. S, "A Machine Learning-Based Employee Layoff
Prediction System," 2025 3rd International Conference on Artificial Intelligence and
Machine Learning Applications Theme: Healthcare and Internet of Things (AIMLA)
4) V. Manish, Y. Manchala, Y. Vijayalata, S. B. Chopra and K. Y. Reddy, "Optimizing
Resume Parsing Processes by Leveraging Large Language Models," 2024 IEEE
Region 10 Symposium (TENSYMP)
5) N. Gangoda, K. P. Yasantha, C. Sewwandi, N. Induvara, S. Thelijjagoda and N.
Giguruwa, "Resume Ranker: AI-Based Skill Analysis and Skill Matching System,"
2024 Sixth International Conference on Intelligent Computing in Data Sciences
(ICDS)
6) A. Mishra, S. Singh, and R. C. Jisha, ―AI-Powered Model for Intelligent Resume
Recommendation and Feedback,‖ 2024 15th International Conference on Computing,
Communication and Networking Technologies (ICCCNT), IIT Mandi, India, pp. 1-8,
June 2024. doi: 10.1109/ICCCNT61001.2024.10726016.
7) D. Ather, G. Singh, N. Chaudhary, P. Verma, R. Kler and A. Gupta, "Analyzing
Trends, Skills Demand, and Salary Prediction in the AI and ML Job Market," 2024
International Conference on Intelligent & Innovative Practices in Engineering &
Management (IIPEM)
8) Y. Cheng, "Research on Personalized Learning Recommendation Model Based on
User Collaborative Filtering Algorithm," 2024 3rd International Conference on
Robotics, Artificial Intelligence and Intelligent Control (RAIIC)
9) S. K. S, "Machine Learning based Ideal Job Role Fit and Career Recommendation
System," 2023 7th International Conference on Computing Methodologies and
Communication (ICCMC).
10) Y. Liang, Z. Lin, X. Shi and H. Ma, "Fast and Accurate Resume Parsing Method
Based on Multi-Task Learning," 2023(IALP).

29
30