Senior software engineer (17+ yrs, TypeScript/React, streaming & TV platforms) building LLM applications — retrieval, LangGraph orchestration, grounding and evaluation. Toronto.
Currently building
- LoomCV — JD-guided resume evidence retrieval and tailoring: hybrid semantic + keyword retrieval over a structured STAR vault, LangGraph orchestration with human review gates, provider-agnostic model routing. Private during development (real career data); curated public release planned at MVP.
Published
-
Claude Code × Ollama on Apple Silicon — running Claude Code against a locally hosted model (Ollama / Qwen 3.8 27B): the configuration that works, three root causes found by capturing the requests, and what it isn't good for。
-
Claude Code on a local model — unattended workflow — the multi-agent development workflow (strategist / implementer roles, ticket templates, hook-enforced review gates) I use on LoomCV.