I build production AI systems at the intersection of LLMs, financial data, and agentic workflows.
Currently a Data Scientist at Think Analytics, where I ship GenAI and RAG systems used by regulators, banks, and AMCs — not demos, real tools handling real data at scale.
- 🏦 Deep focus on Financial AI — compliance, investment research, credit underwriting, and mutual fund analytics
- 🤖 Building multi-agent pipelines with LangChain, LangGraph, and CrewAI
- 📚 Preparing for CFA Level 1 — bridging the gap between AI engineering and finance
- 🔐 Care deeply about data privacy, audit trails, and responsible AI deployment
- 🎓 B.Tech in Electronics Engineering with Minors in AI/ML — DJSCE Mumbai (8.0 GPA)
| Project | Impact |
|---|---|
| GenAI Compliance Surveillance (Capital Markets Regulator) | Detects unregistered finance influencers & classifies MF ads at scale |
| RAG Compliance Intelligence System (Leading Bank/NBFC) | Indexed national regulatory circulars; cut manual tracking effort significantly |
| Mutual Fund Analytics Dashboards (Multiple AMCs) | Power BI layer across AUM, demographics, geography → +46% business growth, −8% churn |
| LLM Real Estate Workflow (Square Yards) | LangChain-powered inquiry handling → +50% lead conversion |
| CI/CD for AI Apps (Square Yards) | Docker + AWS pipelines → −40% deployment time |
AI / LLM
ML / Data Science
Infrastructure & Backend
Visualization
LangGraph · FastAPI · PostgreSQL · Next.js · Docker
Built a full advisory pipeline with specialized agents for ingestion, quant analysis, investor profiling, fund recommendation, compliance validation, and explainability. Evaluated across 500+ simulated investor profiles; achieved +12% Sharpe ratio improvement over baseline in backtesting.
LangChain · FAISS · FastAPI · OpenAI/HuggingFace · Docker
RAG system over 5K+ trade documents achieving 90%+ citation accuracy for regulatory evidence. Fully auditable with versioned knowledge bases and 65% reduction in manual review time.
Kafka · Spark Structured Streaming · BERT · ONNX · PostgreSQL
Real-time pipeline processing 2.9M+ SMS records at sub-3s latency. Fine-tuned BERT classifier on 19K Indian financial SMS messages — 99.3% accuracy, 0.993 macro F1 across 10 transaction categories.
SentenceTransformers · XGBoost · LLM Fallback · Browser Automation
Autonomous agent that observes, classifies, and reshapes YouTube recommendations. Human-in-the-loop feedback loop with weekly retraining — 92% precision, 70% reduction in irrelevant content within 2 weeks.
Building AI that works in the real world — not just in notebooks.