I’m a software engineer at Google, focused on machine learning performance — making large models run efficiently on TPUs. Day to day that means profiling, compiler and framework internals, kernel optimization, and the systems work that turns raw FLOPs into actual throughput. Over 20 years of building software, with earlier stops in .NET/F# tooling and distributed data systems.

What I cannot code, I do not understand.

This blog is where I write up what I learn along the way: ML performance and profiling, accelerators and compilers, BigQuery and data pipelines, and the occasional functional-programming detour.

A few things I’ve built

• tpu_performance_autoresearch_wiki — a TPU-performance knowledge base wired into an autoresearch optimization engine.
• xprof-mcp — an MCP server for the open-source XProf profiler.
• FSharp.Interactive.Intellisense — IntelliSense for Visual Studio F# Interactive.

Elsewhere

• GitHub: vlasenkoalexey
• LinkedIn: Oleksiy Vlasenko