╳ Research

Active research tracks

Two threads I'm actively working on. Each entry below is a living record — I update it as findings change rather than starting a new page for every increment.

Updated Jun 28, 2026 Inference Engineering Continuous batching, KV-cache reuse, and speculative decoding — full write-up with benchmarks. Inference EngineeringActive Updated Jun 20, 2026 Agent Architecture Retrieval and context systems designed for reliable multi-step agent behavior. Agent ArchitectureActive