Third-year IT student at Dwarkadas J. Sanghvi College of Engineering, Mumbai. I work at the intersection of AI/ML engineering and product thinking — most of my projects are systems that combine LLMs, vector databases, and real-time data pipelines into things people can use.

Outside the technical work: I compete in hackathons and case competitions, follow the Indian equity market closely, and tinker with custom ROMs when I'm curious about what's running underneath.

AI-augmented crisis helpline intelligence layer · SOC1 Hackathon

Real-time operator HUD for crisis helpline management. Built on a five-agent system with a MetaAgent for conflict resolution between agent outputs. Multilingual STT via Sarvam AI. Entirely free-tier stack.

Stack: WebRTC · Groq · Sarvam AI · Leaflet.js · OpenStreetMap · Overpass API
Architecture: 5-agent pipeline — Triage Agent, Risk Agent, Resource Agent, Sentiment Agent, MetaAgent (conflict resolver)

Agentic AI backtesting platform for equity strategies

LangChain-powered agent that interprets natural-language trading strategy descriptions, generates backtesting logic, executes against historical NSE data, and returns performance analytics. Uses Pinecone for strategy memory and retrieval.

Stack: LangChain · Pinecone · Python · FastAPI · NSE data
Key feature: Strategy memory via vector embeddings — retrieve similar past strategies before generating new ones

B2B AI trade matchmaking platform

AI-powered platform that matches exporters and importers using semantic search over trade profiles. Migrated backend from Gemini to Groq (llama-3.3-70b-versatile) with round-robin API key rotation across FastAPI and Flask layers. Multi-namespace RAG search via Pinecone.

Stack: FastAPI · Flask · Groq · Pinecone · React
Notable: Round-robin key rotation across both API layers for rate limit management

Indian stock portfolio backend deployed on Hugging Face Spaces

FastAPI backend for NSE-listed equity portfolios. Simulates day high/low tracking, portfolio valuation, sector breakdowns, and returns analysis. Switched data source from Yahoo Finance to nsepython after IP-ban issues.

Stack: FastAPI · nsepython · Python · Hugging Face Spaces
Live: huggingface.co/spaces · Public API endpoints for NSE data

Autonomous paper digest system running on GitHub Actions

Autonomous agent that queries arXiv, Semantic Scholar, and Papers With Code on a schedule, curates relevant AI/ML papers, and delivers a Telegram digest. Extended with a historical AI curriculum mode — delivers landmark papers in chronological order to build foundational knowledge.

Stack: GitHub Actions · Python · arXiv API · Semantic Scholar API · Telegram Bot API
Modes: Daily digest (recent papers) + Historical curriculum (chronological landmark papers)

Academic project · CV + NLP fusion

Detection system that fuses visual and linguistic signals to classify deepfake media. Combines CNN-based frame analysis with NLP-based transcript inconsistency detection.

Stack: Python · PyTorch · OpenCV · Transformers

AI-powered route optimization

Intelligent routing system with natural language query support. Users describe routing constraints in plain English; the system generates optimized paths.

Stack: Python · Leaflet.js · OpenStreetMap · Groq

File and text sharing tool

Lightweight sharing tool deployed on Hugging Face Spaces for quick file and clipboard text exchange.

Stack: FastAPI · Hugging Face Spaces

VoiceForward       → SOC1 Hackathon submission (active)
StratoTest         → Expanding strategy library + backtesting accuracy
Portfolio Insights → Adding options chain data + F&O analytics

• Indian equity markets — equity research, portfolio construction, fundamental + technical analysis of NSE-listed companies
• Custom ROMs — device modding, kernel-level tinkering, understanding what's running at the OS layer
• Case competitions — business strategy under time pressure

Building something with agentic AI, RAG, or real-time data pipelines? Or just want to talk equity research or custom ROMs?