AI Engineer
I build agent-based software solutions that work in production. Combining LLMs with retrieval-augmented generation, tool use, and memory to create reliable, cost-aware AI systems. I write about agentic AI, orchestration patterns, and what actually works.
A deep dive into Google's TurboQuant algorithm — how PolarQuant and QJL combine to compress LLM key-value caches to 3 bits with zero accuracy loss, 6x memory reduction, and 8x faster attention on H100 GPUs. What it means for agentic AI, local inference, and the memory economics of production LLM systems.
A deep dive into building production-ready AI agents with LLMs, tool use, and memory systems. Learn the patterns and pitfalls from real-world deployments.
A visual and mathematical exploration of the Transformer architecture, from self-attention to multi-head attention, with practical implementation insights.