Hi, I'm

I'm a final-year IT student at KJ Somaiya. I work on using AI to solve real-world problems, and what I care about most is the product around the model: who it's for, and whether it actually makes their day easier.

Open to Jan 2027 internships · US citizen

YOU 0 · 0 ME
Mrunmayi, smiling at a café

Projects

Things I've worked on

All projects
01IIT Bombay research

Multimodal Breast Cancer Classification

Combined handcrafted radiomic features with VLM embeddings for automated BI-RADS assessment of mammograms, at IIT Bombay.

  • Computer Vision
  • VLMs
  • Radiomics
  • PyTorch
0210+ hrs/week saved

Abuse Email Triage

Automated triage and summarization for 1,000+ abuse emails a day at Tata Teleservices, cutting manual work by 95%.

  • NLP
  • Python
  • Azure
  • Summarization
SynthAI brings Jira, Git and team activity into one place
03Datathon, 2nd of 31

SynthAI

One knowledge graph for Jira, Git and team activity, with AI agents that help managers plan work and spot risk early.

  • GraphRAG
  • Neo4j
  • LangChain
  • React
Your crypto wallet
04Won 2 prizes

TapOut

An AI agent that lets you pay a friend with a tap, using NFC and crypto payments over x402. Built at ETHMumbai.

  • React Native
  • Solidity
  • x402
  • AI agents
0540% less manual work

Policy Documents RAG

Knowledge-graph retrieval for policy analysis at OMFYS, with adaptive chunking and Redis caching. Cut manual analysis effort by 40%.

  • RAG
  • Knowledge graphs
  • Redis
  • Hugging Face
A travel guide to Jharkhand
06Smart India Hackathon

Padharo Jharkhand

A tourism platform for Jharkhand with a multilingual text and voice chatbot for homestays and destinations, built for Smart India Hackathon.

  • React
  • Flask
  • MongoDB
  • Gemini
  • Sarvam AI

Experience

Where I've worked

Resume
  1. Research InternIIT Bombay
    Dec 2025 – Mar 2026Mumbai

    Multimodal breast cancer classification using radiomics and VLMs for BI-RADS assessment.

  2. AI/ML InternTata Teleservices
    Jun 2025 – Nov 2025Mumbai

    Automated triage and summarization for 1,000+ daily abuse emails with Python and Azure.

  3. AI InternOMFYS Technologies
    Jan 2025 – May 2025Mumbai

    Built a knowledge-graph RAG system for policy documents, cutting manual analysis effort by 40%.

  4. Software Engineer InternOMFYS Technologies
    Jun 2024 – Oct 2024Mumbai

    Built and deployed survey applications using Oracle APEX and REST APIs.

Research

Research & publications

  1. Detecting and Localising Label Leakage in Benchmark Datasets

    Came out of rebuilding my engagement-prediction model after finding label leakage in it.

    First author · under review
  2. How To Tell Whether a Quantum Head Will Help: A Diagnostic Protocol for Hybrid Quantum-Classical Medical Imaging

    When does a quantum classifier head actually add something over a classical one?

    First author · under review

Personal

Notes & less formal things

TIL · today I learned

View all
  • Sep 2026A simple model can be a great leakage detectorIf a plain linear model predicts your label almost perfectly, the label might just be a formula of your features.
  • Sep 2026How you chunk documents matters a lot in RAGAdaptive chunking plus graph-aware retrieval got us a 60% jump in answer accuracy on policy questions.

Less formal things about me

  • Been to 20 countries (and counting!)
  • Coffee, a lot of it
  • Puzzles & chess, always
  • Skincare, nails, dressing up
  • Planes (though flying scares me)

Links

Elsewhere on the internet

Get in touch

Say hi

I'm always happy to talk about AI, products, internships, or where to get good coffee.

manu.bhavsar3@gmail.com