Technical deep dives, tutorials, and insights on ML Systems, Infrastructure, and Software Engineering.
Guides and workflows for AI-powered development tools.
Most developers use Claude Code at 20% of its potential. These 13 workflows -- from parallel instances to feedback loops -- will change how you ship. Fork the repo, work through the exercises, and build muscle memory by end of day.
A comprehensive, evolving guide to effectively using Opencode for AI-powered coding. Learn tips, tricks, and workflows from daily use that will transform your development workflow.
Deep dives into NL2SQL frameworks, benchmarks, and architectures.
I benchmarked three open-source NL2SQL approaches against the same DuckDB database with 50 gold-standard queries. Vanna AI's RAG approach led at 40% execution accuracy, but all three tools reveal hard truths about the state of text-to-SQL.
A deep technical teardown of three leading open-source NL2SQL frameworks -- tracing the query path from natural language to SQL, comparing RAG vs semantic models vs multi-agent architectures, and identifying what works and what doesn't.
LLM inference, serving frameworks, and distributed systems debugging.
A deep dive into root-causing a multi-prefill disaggregation bug in SGLang's NIXL backend -- from symptoms to a three-component failure spanning Rust, Python, and ZMQ.