All my blogs related to AI, new research, coding agents, and math.
Writing in the Margins: Proposed solution to Lost in the Middle
Needle in haystack issue for long context, the margin workflow, chunked prefill, KV cache, and evaluation results.
Parse and Retrieve Dense Tables Accurately with Tensorlake
Handle messy and complex tables for accurate retrieval using hierarchies, coordinates, and summaries.
Fix Broken Context in RAG with Tensorlake + Chonkie
RAG pipelines break on real-world documents because parsers flatten structure and chunkers chunk blindly.
Build Schema-Enforced Pipelines with Tensorlake + Outlines
Pair Tensorlake with Outlines to keep document pipelines clean: structured inputs in, schema-validated generations out.
Building Reliable AI Financial Agents with Automatic Trace Analysis
Trace every tool call, visualize deviations, and stop hallucinated portfolio changes before they hit production workflows.
Detecting and Fixing Tool Call Errors with limbic-tool-use
Add telemetry to every tool invocation, classify failures, and auto-heal retries so production agents degrade gracefully.
Memory in Agents: What, Why, and How
Define agent memory, understand its architectures, and see how it keeps instructions, traits, and goals alive across conversations.
Why Stateless Agents Fail at Personalization
Diagnose why stateless agents forget preferences, then layer in Mem0-style traces to keep context, taste, and tone consistent.
Hybrid RAG with Tavily
Build a hybrid retriever that weighs embeddings, freshness windows, and web snippets, then persists the good stuff back into memory.
Automating Legal Research with Tavily + Quotient AI
Legal research pipelines need grounding, traceability, and evals—this blueprint shows how we shipped that stack for counsel teams.