Virtual Internship

6.0

Infosys Springboard Virtual Internship 6.0

Completion Report
Batch Number : 8,9,10

Start date : 27/11/2025

Name : Mayank Joshi

Internship Duration: 8 Weeks

1. Project Title
Web Application for Product Price Comparison Across E-Commerce Sites

2. Project Objective

The primary objective of this project was to design and develop an intelligent price
comparison application that simplifies the online shopping process through the use of
Artificial Intelligence and automation.

In the current digital landscape, online shopping has become ubiquitous, yet consumers
face a significant challenge: finding the best price for a specific product across a
fragmented market. A single item is often listed on multiple platforms—such as Amazon,
Flipkart, eBay, and Walmart—with varying prices, discounts, and shipping fees.
Manually navigating these sites to compare costs is inefficient and time-consuming.

It aims to eliminate this manual effort by offering a centralized solution where users can
simply upload an image of the product they wish to purchase. The system is engineered
to automatically identify the product using a deep learning model and instantly search
multiple online stores to locate the lowest available price.

Key Project Goals:

 Automate Price Discovery: To eliminate the need for manual searching by

automating the comparison process across rival e-commerce platforms.
 Ensure Cost-Effectiveness: To empower users to make informed, cost-effective
purchasing decisions by presenting the lowest available prices and best deals.
 Enhance Efficiency: To significantly reduce the time users spend searching for
products online by providing instant, accurate results.
 Improve User Experience: To deliver a user-friendly, intelligent shopping assistant
that provides accurate product identification and detailed information in a single
interface.

pg. 1
Virtual Internship
6.0
Ultimately, this project aims to demonstrate the real-world application of AI in the e-
commerce domain, providing a scalable solution that bridges the gap between visual
search technology and practical consumer needs.

pg. 2
Virtual Internship
6.0

3. Project description in detail

It is an AI-powered price comparison application designed to simplify and enhance the

online shopping experience by enabling users to find the best available price for a product
using just an image. This innovative tool bridges the gap between physical interest and
digital purchasing, acting as a personal shopping assistant that eliminates manual data
entry.

Problem Statement: Consumers often face the challenge of manually searching for the
same product across multiple websites to find the best deal. This process is time-
consuming, inefficient, and sometimes confusing due to variations in product names and
descriptions. Additionally, a single product may be listed with different shipping costs,
discount structures, and seller ratings across platforms like Amazon and Flipkart, making
it difficult to calculate the true final cost without checking each site individually.

Solution & Technical Approach: The Compario application solves this problem by
allowing users to upload an image of the product they wish to purchase, replacing text-
based search with an intuitive visual lookup.

 Preprocessing: Once the image is uploaded, the system performs image

preprocessing techniques such as resizing, noise reduction, and enhancement to
improve recognition accuracy. Specific techniques like pixel value normalization and
contrast enhancement are applied using OpenCV to ensure the model remains robust
against poor lighting or low-quality uploads.
 AI Recognition: A deep learning-based Convolutional Neural Network (CNN)
model is used to analyze the image and identify the product. This model was trained
on a comprehensive dataset of product images to effectively detect categories and
brands, predicting the most relevant product match to trigger the search process.
 Data Collection: After successful recognition, the system searches multiple e-
commerce platforms (Amazon, eBay, Walmart, Flipkart, etc.) using APIs or web
scraping techniques. Python scripts are utilized to clean and standardize this incoming
data, ensuring that product titles, ratings, and availability statuses are consistent
regardless of the source platform.
 Comparison Algorithm: Collected prices are processed to identify the lowest
available price, including seller details and availability. This algorithm also
intelligently handles variations in currency formats and calculates additional shipping
charges to ensure a fair and accurate comparison for the user.
 User Interface: The application provides a clean dashboard displaying the product
image, description, and a list of prices, with the best deal highlighted. The interface is
built using responsive web technologies (HTML/CSS/JS), allowing users to easily
view price trends and make quick, informed purchasing decisions from any device.

pg. 3
Virtual Internship
6.0
4. Timeline Overview
Week Activities Planned Activities Completed
Week 1 Project setup, repository Environment setup,
creation, and GitHub repo
requirements gathering initialization, and
defined project scope.
Week 2 Image upload Implemented Flask
module and file upload handling
preprocessing and image validation
logic scripts.
Week 3 AI Model integration Integrated AI processor
(CNN/Deep Learning) for image recognition
and feature extraction.
Week 4 Web scraping strategy Developed logic to fetch
and API integration and aggregate pricing
data from simulated
datasets.
Week 5 UI Development and Created responsive
frontend framework templates for login,
setup search, and product
display pages.
Week 6 Product display and Connected backend
details integration search logic to frontend
results, enabling visual
matches.
Week 7 Testing, refinement, and Conducted model
bug fixing accuracy tests and fixed
session management
issues.
Week 8 Documentation and Completed final project
Final Review report, code cleanup,
and presentation
preparation.

5a. Key Milestones

Milestone Description Date Achieved
Project Kickoff Finalized scope, role Week 1
allocation, and tool
installation
Prototype/First Draft Functional image upload Week 3
system with
preprocessing and basic
UI
AI Integration Successful integration of Week 5
the Deep Learning model
for product recognition
Prototype Working prototype Week 8
capable of taking an
image and outputting
product matches
Final Submission Final demo and Week 8
presentation

pg. 4
 Virtual Internship
6.0

5b. Project execution details

The project was executed using a modular architecture to ensure scalability and
maintainability. The system was divided into distinct functional components that interact
seamlessly:

 Authentication Module:
o Security: Implemented secure user registration and login functionalities using
the [Link] library for password hashing and verification,
ensuring sensitive user credentials are never stored in plain text.
o Session Management: Utilized Flask sessions to maintain user state across
requests, tracking critical user data such as search history, specific delivery
city preferences (e.g., mapping users to cities like Mumbai or Delhi), and role-
based access control (User vs. Admin).
o Route Protection: Custom decorators like @login_required and
@admin_required were implemented to secure specific endpoints and prevent
unauthorized access to sensitive pages.
 Image Processing Pipeline:
o Secure Uploads: The system supports various image formats (JPG, PNG,
WEBP). Upon upload, files are validated for security and renamed using
uuid.uuid4() to generate unique, collision-free filenames before being saved
to the server's static/uploads directory.
o Preprocessing: Before analysis, raw images undergo a preprocessing stage
using OpenCV techniques. This includes resizing images to standard
resolutions, noise removal, and pixel value normalization to ensure the input
data is consistent for the AI model .
o AI Prediction: The core intelligence is handled by the ai_processor
module. It utilizes a pre-loaded Deep Learning model (integrated via
load_ai_model) to analyze the visual features of the uploaded image and
predict the product category or generate relevant search tags.
 Search & Comparison Logic:
o Database Search: The system maintains an in-memory PRODUCT_DATABASE
loaded from CSV datasets. When a prediction is made, the
find_product_matches function queries this database to find relevant items.
o Smart Filtering: The logic applies dynamic filters based on the user's session
data. It filters results not just by product similarity, but also by delivery
availability in the user's selected city and price ranges.
o Price Aggregation: The system aggregates pricing data from multiple
simulated store sources (CSV files) to present a comparative view, identifying
the global best price among all matches.
 Admin Features:
o Dashboard Analytics: A dedicated admin interface was developed to
visualize system health, displaying real-time statistics such as total registered
users, active sessions, and total search volume.

pg. 5
Virtual Internship
6.0
o Data Management: Administrators have the capability to dynamically
refresh the product database using the reload_csv function, allowing for
inventory updates without restarting the application.
o Activity Logging: The system maintains a comprehensive log file
([Link]) that records user actions—such as logins, searches, and city
changes—which the admin can monitor for security and usage patterns.

pg. 6
Virtual Internship
6.0
6. Snapshots / Screenshots

pg. 7
Virtual Internship
6.0

pg. 8
Virtual Internship
6.0

pg. 9
Virtual Internship
6.0

pg. 10
Virtual Internship
6.0

pg. 11
Virtual Internship
6.0

7. Challenges Faced

 Image Quality Issues: Low-quality or blurry user uploads initially led to poor
recognition accuracy. We addressed this by implementing robust preprocessing
techniques like normalization and resizing.

 Overfitting: The AI model initially struggled with overfitting on training data. We

mitigated this by diversifying the testing dataset with multiple real-world images.

 Data Consistency: Handling CSV data for product comparisons required careful
parsing to ensure prices and delivery times were accurate across different "stores".

8. Learnings & Skills Acquired

Advanced Full Stack Engineering:

 Gained expert-level proficiency in Python Flask development, moving beyond

basic routing to implement secure session management, custom decorators (e.g.,
@login_required), and cryptographic password hashing using
[Link].
 Mastered server-side rendering using the Jinja2 templating engine to dynamically
inject Python data into HTML interfaces, creating a seamless user experience
between the backend logic and the frontend presentation.

Applied AI & Computer Vision Pipelines:

pg. 12
Virtual Internship
6.0
 Learned the practical implementation of Convolutional Neural Networks
(CNNs), specifically focusing on the inference pipeline rather than just model
theory.
 Developed robust image preprocessing workflows using OpenCV, implementing
techniques like pixel normalization, noise reduction, and resizing to standardize
user inputs before they reach the classifier, significantly reducing model inference
errors .
 Solved the challenge of integrating heavy Machine Learning models into a
lightweight web framework, managing memory usage and latency during real-
time predictions.

Scalable System Architecture:

 Designed a modular application architecture that strictly separates concerns

between Authentication, Data Handling (utils.csv_loader), and AI Processing
(utils.ai_processor).
 Architected a custom file-based database system using CSVs to simulate a
relational database, writing optimized parsing logic to handle data aggregation
from multiple "store" sources efficiently.

Version Control & DevOps Practices:

 Improved proficiency in GitHub for version control, managing branch strategies

for different modules (Frontend vs. AI) to minimize merge conflicts.
 Implemented strict file validation and security protocols for user uploads (UUID
renaming) to prevent server-side vulnerabilities, simulating real-world DevOps
security standards.

Team Leadership & Project Management:

 Strategic Planning: As the Team Lead, I orchestrated the entire development

lifecycle, breaking down the 8-week timeline into actionable sprints and assigning
specific modules (UI, AI, Backend) based on individual team member strengths.
 Technical Mentorship: Acted as the primary technical point of contact, guiding
team members through complex challenges such as API integration and model
overfitting, ensuring the project met its technical milestones on time.
 Collaboration Management: Facilitated seamless communication between the
frontend and backend developers, ensuring that the data contracts (JSON
structures) matched the UI requirements, resulting in a successful final
integration.

9. Testimonials from team

The internship project on the "Compario - AI Powered Price Comparison

Application" served as a highly enriching and rewarding professional milestone for
pg. 13
Virtual Internship
6.0
our entire team. It provided us with a unique opportunity to bridge the gap between
academic theory and practical implementation, allowing us to apply complex concepts
in Artificial Intelligence, image processing, and full-stack web development to solve a
tangible real-world problem. Beyond the technical coding, building this AI-powered
system from scratch offered us deep insights into the intricacies of the complete
software development lifecycle (SDLC), teaching us how modern software
applications are architected, developed, tested, and deployed in industry-standard
environments.

Furthermore, the collaborative nature of this project significantly enhanced our

professional competencies, particularly in team communication, strategic planning, and
adaptive problem-solving. By working cohesively where each member took ownership
of specific modules—ranging from the AI backend and data collection logic to the
responsive user interface—we were able to successfully integrate diverse technical
components into a seamless, fully functional application. This comprehensive
experience, supported by invaluable mentorship, has significantly strengthened our
technical foundation and boosted our confidence, leaving us thoroughly prepared to
tackle complex challenges in future professional roles within the software and AI
industries.

pg. 14
Virtual Internship
6.0
10. Conclusion
Bringing Compario to life was about more than just fulfilling an internship requirement; it was
about solving a genuine frustration we all face as consumers. We’ve all been there—switching
between five different tabs trying to figure out if we’re actually getting a good deal. With this
project, we managed to turn that tedious process into something simple: snap a picture, and
let the code do the heavy lifting. Seeing our application successfully take a raw image and
instantly return price comparisons felt like a massive win, proving that AI isn't just a buzzword,
but a practical tool that can make everyday life a little bit easier.
Under the hood, this project forced us to grow as engineers. It wasn't enough to just write a
few scripts; we had to build a robust system where a complex AI model could "talk" to a web
backend seamlessly. We spent days refining the Convolutional Neural Network (CNN) to
ensure it didn't just guess randomly, and we worked hard to ensure the interface was
responsive and clean. Integrating these distinct pieces—from the image processing pipeline
to the price fetching algorithms—taught us the importance of modular architecture and
rigorous testing in a way that no classroom lecture ever could.
The journey was filled with its fair share of hurdles, which ultimately became our best
teachers. We wrestled with real-world issues, like how a slight change in lighting could
confuse our model or how different websites format their prices completely differently .
overcoming these challenges—by implementing smarter preprocessing and robust data
handling—gave us the kind of hands-on problem-solving experience that is essential for a
career in software development. It bridged the gap between theoretical computer science and
the messy reality of building live applications.
Looking back, we are incredibly proud of what Compario has become—a scalable, intelligent
assistant that actually works. While it is a fully functional product today, we already see so
much potential for where it could go next, from price drop alerts to a mobile app version. This
internship didn't just improve our coding skills; it gave us the confidence to lead projects and
design systems from the ground up, leaving us ready and excited to step into the professional
tech world.

11. Acknowledgements
We sincerely express our gratitude to Infosys Springboard for providing us with
this valuable learning opportunity through the Virtual Internship program. This
internship has helped us gain practical exposure to real-world applications of
artificial intelligence and software development.

We would like to extend our heartfelt thanks to our mentor Berlin Mam for her
continuous guidance, technical support, and encouragement throughout the
project. Her valuable insights and timely feedback played a crucial role in the
successful completion of this project.

We are also grateful to our coordinator Mahalakshmi Mam for her constant support,
motivation, and effective coordination during the internship period. Her efforts
ensured a smooth and well- organized learning experience.

Finally, we thank all our team members for their cooperation, dedication, and
teamwork, which made this project a success.

pg. 15