Michael Rakutko

Python, SQL, dbt for the analytics core. LLM eval pipelines, local ASR, and data plumbing for AI tooling. Deep in analytics, wide across the stack - from warehouse models to agent harnesses.

Building reliable eval harnesses for coding agents - measuring what actually breaks in agentic workflows. Digging into local ASR & LLMs and how far small models close the gap with cloud APIs.

Analytics for AI coding agents: turning agent traces into signal about adoption, ROI, and output quality. Plus an eval harness that catches regressions before they reach main.

Comparing notes on LLM evals, AI-coding-agent adoption metrics, and analytics for eng orgs. Happy to talk shop with anyone measuring how teams actually use AI.