I build production systems — not just projects. I care about clean architecture, developer tooling, and shipping things.
Currently: Building AI-powered financial document systems · Seeking SWE and AI internships
These are my most substantial, production-grade projects — the ones I'd walk you through in an interview.
Fincite — AI-Powered Financial Filing Research Assistant
Evidence-grounded RAG system for querying Indian listed companies’ annual reports and investor presentations
Building an end-to-end financial research assistant that answers natural-language questions over company filings with verified financial values, page-level source citations, and confidence scores.
Designed a multi-agent LangGraph workflow that separates query planning, document retrieval, financial metric extraction, verification, and response synthesis. Includes a verification-and-retry mechanism that cross-checks extracted values against source statements and automatically re-runs retrieval when inconsistencies are detected.
Python LangGraph FastAPI OpenAI / Qwen PostgreSQL pgvector React Docker
BLE Location Tracker — AirTag-Style IoT Tracking Platform
End-to-end ESP32-C3 BLE tracker with a Web Bluetooth PWA, GPS relay reporting, and an Express/SQLite backend
Designed and built an end-to-end BLE location tracking system that pairs a custom ESP32-C3 tracker with a browser-based Progressive Web App and a Node.js backend — covering firmware, wireless protocol design, and the API layer in one project.
The ESP32 exposes a custom BLE GATT service with characteristics for tracker identity, device status, battery level, and alarm control, handling reconnection, notification publishing, and auto-cancelling alarms. The PWA discovers nearby trackers with the Web Bluetooth API, reads GATT characteristics, subscribes to status notifications, sends PING and alarm commands, captures the relay device's GPS coordinates, and reports sightings to an authenticated Express API that persists location history in indexed SQLite tables and renders live movement trails on a Leaflet map.
Extended the project with service-worker caching for installable PWA support, JWT-based tracker ownership, and Python experiments using Bleak that apply a moving-average filter plus a path-loss model to estimate distance from noisy RSSI readings.
ESP32-C3 C++ / Arduino BLE Web Bluetooth API JavaScript PWA Node.js Express SQLite JWT Leaflet.js Python
Languages
Frontend
Backend & Databases
IoT & Embedded
AI & LLM
DevOps & Tools
Data Engineering
Open to AI internship opportunities.
