B.Tech Information Technology undergraduate at VESIT, Mumbai (9.56 CGPA), holding student leadership roles at ISTE-VESIT and QuestIT. I enjoy building production-ready AI systems that combine modern LLM workflows, explainability, and full-stack engineering.
My interests include AI Agents, Distributed Systems, Explainable AI, and Developer Infrastructure.
|
Explainable AI governance for discretionary decisions, built on IBM Granite / watsonx.ai. Surfaces undefined terms in governing text (e.g. FIFA's "deliberately"), runs a 4-persona interpretive engine, and generates an exportable Discretion Disclosure Report. Stack: React 19, Vite, TypeScript, Vercel Serverless, IBM watsonx.ai (Granite) Built for: IBM SkillsBuild AI Builders Challenge, June 2026 |
Mule-account detection for banking fraud pipelines using XGBoost with SHAP-based explainability, built so flagged accounts come with a defensible, auditable reason rather than a black-box score. Stack: Python, XGBoost, SHAP Built for: PSBs Cybersecurity, Fraud & AI Hackathon 2026 (Bank of India × IIT Hyderabad) |
|
Civic accountability platform for surfacing and tracking local governance issues, orchestrated through a five-agent Gemini 2.0 Flash pipeline — separate agents handle intake, verification, reasoning, and reporting rather than one monolithic prompt. Built for: VIBE2SHIP Hackathon (Coding Ninjas × Google for Developers), PS2 — Community Hero, June 2026 |
Multi-agent system for detecting contradictions across enterprise policy documents — the core idea being that siloed document review misses conflicts that only show up when clauses are compared across documents. Built for: Microsoft Agents League AISF 2026 Hackathon |
|
Voice-first AI assistant bridging the digital-education divide — lets users query educational content in natural language instead of navigating text-heavy interfaces. Stack: React, Node.js, Google Gemini, Web Speech API |
|
