This repo collects technical notes around systems I am building and studying: long-horizon agent workflows, RAG/document intelligence, payout correctness, retrieval architecture, evaluation, and product engineering.
The goal is to keep a public trail of how I think about applied AI and product systems. These notes are not polished research papers. They are working documents that connect implementation details, system-design tradeoffs, and the projects I am building in public.
- Agent State vs Memory
- Payout System Correctness
- Vision-RAG Notes
- RAG Evaluation Notes
- Knowledge Graphs for Retrieval
- Open Source Contribution Plan
- Open Source Targets
- Durable state for long-horizon agent workflows.
- Retrieval and document intelligence for layout-heavy files.
- Backend correctness in money-movement simulations.
- Evaluation loops for RAG systems.
- Practical open-source contribution habits.
- Stateframe - file-first task-state ledger for agent workflows.
- PayRail - payout correctness demo with ledger accounting and idempotency.
- Portfolio - public project surface for applied AI and product engineering work.
Active notes repo. Content will evolve as projects, experiments, and contribution work become more concrete.