Hi, I'm Aman - AI Engineer.
About me.
I build applied AI and product systems for messy information workflows: RAG, document intelligence, payout correctness, retrieval evaluation, and backend systems that stay inspectable beyond the demo.
Skills & Stack
AI & Systems
Infrastructure & Tools
Pune, India
(Asia/Kolkata)
Experience
Worked on document intelligence and retrieval workflows for layout-heavy PDFs, knowledge graph extraction, and practical RAG evaluation.
Featured Projects
Strongest proof-of-work across AI systems, payout correctness, product infrastructure, and technical writing.
GitHub Activity
Open Source Contributions
Building a contribution trail through small, inspectable PRs.
Technical Notes
Notes that connect the projects to the systems thinking behind them.
Agent State vs Memory
A note on why long-horizon agents need durable state in addition to memory, traces, and orchestration.
Payout System Correctness
PayRail as a concrete model for idempotency, append-only ledger accounting, row locks, retries, and double-spend prevention.
Vision-RAG Notes
Where OCR-first RAG breaks down, why page-level retrieval matters, and how visual retrieval changes document AI workflows.
RAG Evaluation Notes
A practical frame for separating retrieval quality, generation quality, groundedness, latency, cost, and failure analysis.
Knowledge Graphs for Retrieval
When graph traversal helps RAG, how it complements vector search, and where hybrid retrieval becomes useful.