AI systems and full-stack developer. I build things that actually work , retrieval pipelines, LLM infrastructure, on-chain applications, browser-level security tooling. Prefer understanding systems from the inside out rather than wrapping APIs.
| Project | What it does | |
|---|---|---|
RAG |
Tokaroo | Adaptive RAG evaluator - retrieval failure analysis, BM25 + cross-encoder reranking, context compression |
LLM |
Exora | Real-time competitive intelligence engine - multi-LLM pipelines, SSE streaming, BYOK architecture |
ML |
URLGuard | On-device phishing detector - TF.js model, 16 lexical features, homograph detection, 88.8% accuracy · 13th / 925 at Amplicode |
ML |
ChunkdUp | Chunking strategy explorer - how LLMs process and retrieve context, without abstraction layers |
Web3 |
SplitMate | On-chain expense splitting - Solana smart contracts, Socket.io real-time sync, AI validation |
Rust |
Shinobi | High-performance CLI for Solana NFT and wallet forensics |
Primary focus: AI infrastructure · retrieval systems · LLM pipelines · backend
Open to internships: AI infra · backend · DevOps
[email protected] · @AdityaPat_
Curious by default. I like understanding how things actually work.



