Builder system

Agentic development workflows that turn AI tools into operating leverage.

My GitHub surface is strongest when it shows the operating system behind the products: MCP servers, LLM observability, evaluation frameworks, CLI tools, native apps, and release workflows.

MCP and CLI

Trackly CLI and MCP server let AI assistants work with recruiting data instead of only showing it in a UI.

Observability

Langfuse, PostHog, Datadog-style workflows, and structured review gates make AI systems inspectable.

Evals

Golden test sets, classification evaluation, and review loops keep automation from becoming invisible risk.

Representative public repos

trackly-cli: CLI and MCP server.
llm-observability: self-hosted observability patterns.
umami-mcp-server: Go MCP server for behavioral analytics.
marketplace-refund-policy-kit: policy simulation and shadow-mode evaluator.