Team Members:Ritika M Patil[4BD23CS171], Ruthu R B[4BD23CS173], Syed
Usaid[4BD23CS227], Vikas K R[4BD23CS243]
Project Title:
Smart autonomous grocery assistant for the elderly:
The system generates shopping lists from spoken input and analyzes past purchase history
stored in the app to suggest frequently bought or budget-friendly items. It also recommends
healthy alternatives based on user preferences and allows exporting the final list for online
grocery ordering.
Problem Statement:
Smart grocery assistant for elders in which the system generates/creates shopping list by taking
input orally/verbally by the one who is using this assistant to create shopping list, it also suggests
some items by analyzing ur previous purchases which would also be budget friendly and also
recommends healthier alternatives and the final list will be exported to for the online grocery
ordering.
Team Members : Pavan Kumar Hiremath [4BD23CS129], Renu K G [4BD23CS169],
Sachin M H[4BD23CS175], Sakshi P B [4BD24CS413]
Project Title:
Goal-oriented Virtual Study Assistant
An AI agent that helps students plan study goals, reminds them, adapts based on
performance, and even explains concepts using retrieval-augmented generation.
Technologies: Langchain Agents, LLM(GPT4), Pinecone/FAISS, Streamlit.
Problem Statement;
Many students procrastinate or lose track of their academic goals due to distractions
and poor planning. This results in last-minute cramming, high stress, and lower
retention of knowledge. An Al-powered study assistant can continuously track goals,
send reminders, and adapt study plans dynamically to keep students on track.
Team Mates: Pragna S M [4BD23CS137], Pruthvi Raj R C [4BD23CS151],
Valmiki Udaya Deepika [4BD23CS233] & Siddarth K Kathare [4BD24CS418]
Project Title: Agentic AI Student Peer Feedback Portal.
Design and develop an Agentic Al Student Peer Feedback Portal featuring sentiment
filtering.
Problem Statement:
Peer feedback can help improve learning, but unclear or rude comments make it less
[Link] Agentic AI-powered portal lets students share feedback on classes, labs,
and campus life without revealing their names .It blocks harmful content, checks the
tone, and creates short reports to help lectures and staff make better decisions
unclearly useless.
Team Members: Sanjana Gowda (4BD23CS187) , Rakshitha D S (4BD23CS161) ,
Vinayaka P J (4BD23CS247) , Prajwal S (4BD23CS147)
Project Title: Student Performance Tracker with Machine Learning and AI
To build a system that tracks and predicts student academic performance using AI/ML techniques,
offering insights to educators, students, and parents for timely interventions.
Problem Statement: The Intervention Mismatch
Current predictive models are becoming proficient at flagging that a student is at risk, but they
operate as a diagnostic black hole, failing to identify why. This leads to a critical intervention
mismatch, where a system might recommend more practice problems to a student whose real issue
is test anxiety or a lack of foundational knowledge. The core problem is that these systems treat all
failure signals as equal, offering generic solutions that are ineffective for the complex, individual
reasons a student truly struggles.
TEAMMATES: Sania S(4BD23CS183),Sharanya S B (4BD23CS195),
Vivek V(4BD23CS249),Shwetha Y B(4BD24CS205)
PROJECT TITLE:Intelligent Multi-Agent Campus Guide(Virtual)
A virtual, multi-agent system that autonomously assists students and visitors by
providing campus navigation, event updates, and department-specific help. Each
agent specializes in tasks like routing, event notifications, or FAQs, working
together to create a seamless virtual campus experience.
PROBLEM STATEMENT:students and visitors often face difficulties navigating
campuses, finding event updates, and accessing department-specific information.
Existing solutions are static and lack real-time assistance. A virtual multi-agent
system can address these issues by autonomously providing navigation, event
alerts, and FAQs for a seamless campus experience.
TEAM MEMBERS:
SRUJAN H.S(4BD23CS215),RAKESH K.U(4BD23CS157),SRUSHTI S.M(4BD23CS219)
NANDINI R(4BD24CS408).
PROJECT TITLE:
Federated Learning-Based Spam Detection on Mobile Devices.
Use FL to collaboratively train a spam classification model across multiple user’s email or
SMS datasets without accessing personal messages.
PROBLEM STATEMENT:
Privacy-Preserving Spam Detection.
Mobile devices receive a high volume of spam messages, but existing centralized
detection systems require uploading personal message data to servers, raising
privacy concerns. There is a need for a privacy-preserving spam detection method
that keeps raw data on-device while still enabling global model improvement.
Team Members: Poojitha P[4BD23CS133], Ramesh H M[4BD23CS163],
Vinay V Divigihalli [4BD23CS245] & Sindhu Shree GN[4BD23CS207]
Project Title: Automated meeting summarization with the conceptual
prompts
Automated meeting summarization leverages AI to condense lengthy meeting
recordings into concise, actionable summaries.
Problem Statement:
Meetings often generate lengthy discussions that are difficult to review and
summarize manually. There is a need for an AI-based system that can automatically
generate concise, accurate summaries from meeting recordings or transcripts using
conceptual prompts, helping users quickly grasp key points and action items.
Team Members: Pragathi M K[4BD23CS135], Ramya M M[4BD23CS165],
Srushti Bahusaheb Jambagi[4BD23CS217], Srujan C A[4BD24CS419]
Project Title: XAI-Enabled Health & Fitness Recommendation System
Problem Statement:
Most fitness apps give generic diet and exercise plans without explaining the reasoning
behind them, reducing user trust and personalization. There is a need for a system that
uses Explainable AI (XAI) to generate personalized recommendations from health data
and clearly justify each suggestion, improving transparency, engagement, and health
outcomes.
Team Members: Shilpa Doddappa Tallur [4BD23CS189],
Sanjana T [4BD23CS],
Prajwal V Pawar [4BD23CS]
Project Title:
Self Supervised Learning for Few Shot Visual Classification
Training a model to classify visual categories using minimal labelled data, leveraging self-
supervised techniques such as contrastive learning or masked image modelling.
Problem Statement:
Image classification is widely used in healthcare, security, and environmental monitoring.
Traditional models require large amounts of labeled data to perform well. In many real-world
situations, especially for rare or specialized categories, collecting such labeled data is costly
and time-consuming. This lack of labeled data limits model accuracy and usefulness. There is
a need for methods that can work effectively with minimal labeled data in resource -
constrained settings.