Netflix's engineering culture taught me how to think about software. Now I build AI tools and apps that work for real people — not just in a notebook.
Turning ideas into intelligent systems
I've always been more interested in why things work than just making them work.
That curiosity has slowly turned into building — systems, ideas, and experiments that try to answer small but interesting questions.
I don't see projects as checkboxes. I see them as iterations of thinking — each one a little more refined than the last.
Right now, I'm especially curious about how intelligent systems interact with real users — and where they fail in unexpected ways.
The interesting part isn't when systems work — it's when they don't, and why.
Where I've worked
Research Assistant
Rutgers University — Edward J. Bloustein School of Planning and Public Policy
Part-time- Conduct research with Prof. Jim Samuel on NLP and agentic AI, developing agents for ethical decision-making.
- Member of the organizing committee for the RAISE-26 (AI – Informatics – Data Science) Competition.
- NLP
- Agentic AI
- Research
Part-Time Lecturer
Rutgers University — Department of Computer Science
Part-time- Serve as part-time lecturer for CS 206: Introduction to Discrete Structures II under Prof. Samaneh Gholizadeh Hamidi, supporting 58 students.
- Lead weekly recitation sessions covering graph theory, combinatorics, and discrete mathematics.
- Hold regular office hours to reinforce core concepts and provide individualized student support.
- Discrete Math
- Graph Theory
- Teaching
Research Intern
IIT Bombay — Industrial Engineering & Operations Research
Research- Engineered a multi-user Excel interface using VBA with MongoDB integration for collaborative data management.
- Integrated a Reinforcement Learning model for the Vehicle Routing Problem with a scalable FastAPI backend.
- Built real-time monitoring dashboards via Highcharts for live operational insights.
- FastAPI
- MongoDB
- VBA
Full Stack Developer Intern
Pack8
Internship- Developed a dynamic web dashboard using React.js and PostgreSQL, replacing the company's Excel-based system and enhancing data visualization and analysis capabilities.
- Implemented backend services with Node.js, providing real-time data handling and improved customization options.
- React.js
- Node.js
- PostgreSQL
Things I've built
Multi-Agent Systems & LLM
AI Personal Assistant
Hierarchical multi-agent system using LangGraph and LangChain to manage emails, calendar, to-do lists, and web research via WhatsApp. Supervisor agent orchestrates specialized sub-agents.
- LangGraph
- LangChain
- Multi-Agent
- WhatsApp API
- Python
NLP & Browser Extension
Tweet Mention Recommendation
Custom BERT model achieving 81.56% accuracy classifying civic issues to automate department tagging. Browser plugin + Streamlit portal with Llama 2. Published at IEEE ISC2-2024.
- BERT
- Llama 2
- JavaScript
- Streamlit
- MongoDB
Applied AI & Data Pipelines
News Aggregator AI Agent
Scalable distributed news aggregator with LLM-powered summarization, keyword extraction, and topic clustering. Fetches from multiple sources in real time with batch processing.
- LLMs
- Distributed Systems
- Python
- NLP
Productivity App
Balance Desk
Full-stack productivity and balance tracking application to help users manage daily tasks, habits, and wellness routines effectively.
- React.js
- Node.js
- MongoDB
Finance App
Expense Tracker
Personal expense tracker with category-based analysis, visual spending breakdowns, and budget management to keep finances in check.
- React.js
- JavaScript
- CSS
Unsupervised ML & Remote Sensing
Aerosol Retrieval & Classification
Two-stage ML algorithm to retrieve aerosol properties (AOD, AE, FMF, FAOD) using MODIS satellite data. Clusters aerosols into 9 types. Submitted to Journal of the Indian Society of Remote Sensing.
- Scikit-learn
- Clustering
- Python