Building things that actually solve problems — AI, computer vision and automation, engineered end-to-end from data pipeline to interface.
2+ yrs in production Winner · Nirman 2026 — ₹75K prize
I build AI systems that actually ship — owning every layer from data pipeline and ML model to API and interface. Most of them built in days, not months.
I’m a full-stack developer and AI/ML engineer with 2+ years of production experience. I’ve shipped hybrid-retrieval RAG engines, computer-vision pipelines, fraud-detection models, React dashboards and Flutter apps — comfortable owning a feature from raw data all the way to a polished UI.
Currently in my B.Tech in Computer Engineering (2023–2027) and the winner of the Nirman 2026 hackathon at Amity University Mumbai. I like problems that are hard, deadlines that are tight, and software that earns its place.
AI-Powered BIS Standard Recommender
A hybrid-retrieval RAG engine over a 929-page regulation corpus. PDF ingestion → 559 structured records → BM25 + dense retrieval → Reciprocal Rank Fusion → cross-encoder rerank → an IS-code whitelist guard that makes hallucinated codes structurally impossible.
Smart Infrastructure Command Center
A three-surface civic ecosystem — a government dashboard, a citizen PWA and a zero-install WhatsApp bot reaching 500M Indian users. A triple-model CV pipeline detects road damage fully offline at 757ms/image, while a monsoon-adjusted engine predicts failure timelines.
FinTech Fraud Detection System
An end-to-end fraud-detection pipeline over 100K+ unlabelled transactions. Five stages — cleaning, EDA, 12-feature engineering, an Isolation Forest + XGBoost ensemble with SMOTE, and explainability — feeding a React dashboard with seven analytical tabs.
Voice-Driven Mock Interview Platform
Feed it a job description and a résumé; it generates role-tailored interview questions you answer by voice or by code. Groq / Llama 3.3 70B scores answers in real time and returns a week-by-week study roadmap with full session playback.
Group Motorcycle Riding App
A mobile app for Indian biker groups — live group tracking, turn-by-turn nav and 1×–10× ride replay. A full safety stack adds one-tap SOS, accelerometer crash detection, bike-tilt fall detection and fuel-range estimation across 100+ models, with bilingual English/Hindi voice commands.
Premium Home Bakery Website
A cinematic Next.js marketing site — a word-by-word animated hero, parallax desserts, Framer Motion scroll-reveals, magnetic buttons and Lenis smooth scrolling. A cart drawer composes a pre-filled WhatsApp order, so the whole shop runs with zero backend.
Built CrackWatch in 48 hours — a three-surface civic infrastructure platform spanning a government dashboard, a citizen PWA and a WhatsApp bot.
₹75,000 grand prizeBuilt BIS Compass single-handed in 7 days — a hybrid-retrieval RAG engine that beat all three rulebook targets by 2×–11×.
Built FraudShield — an end-to-end fraud-detection pipeline and analytical dashboard over 100K+ transactions, finishing among the top 5 teams.
Shipped responsive React + TypeScript web apps with REST integrations. Proposed performance improvements, owned cross-browser testing and bug-fixing, and turned client requirements into shipped features under tight timelines.
Currently pursuing my bachelor’s degree — alongside HSC at Matoshree Prabodhini Junior College (2021–2023).