writing / notes

Notes that make the research program legible.

A strong AI researcher portfolio needs more than a list of publications. This page is the planned surface for technical essays, research notes, and design memos that explain how the artifacts fit together.

planned essay

Formalized research memory for mathematical agents

Why a domain like coding theory needs formal statements, proof trajectories, graph memory, and MCP-style access before a theorem-proving agent can be useful.

LeanAI4Mathresearch memory
revision note

Benchmarks as auditable processes, not files

A public-facing explanation of the EntropyMath / EntropyMaLean direction: lineage, validation contracts, solver traces, and what Lean does and does not guarantee.

EntropyMathevaluationprovenance
field note

From RAG to agent-readable knowledge substrates

What SOGAMBOT, MindBuddhi, regulatory RAG, and TaxCanvas-style retrieval taught me about source-grounded assistants beyond generic chatbot UX.

RAGagentsdeployed systems

First goal: publish three source-backed research notes.

The notes should be short enough to read, concrete enough to cite, and connected to public artifacts rather than generic AI commentary.

Back to research