Thorin Tabor

He / Him

Thorin Tabor

Staff Software Engineer  ·  Agentic AI, MCP & RAG

Nineteen years of full-stack engineering. Open source developer. Cautiously optimistic about agents.

thorin@tmtabor.io LinkedIn Bluesky GitHub

Recreationally, I’ve been coding since grade school. Professionally, I have over nineteen years of experience as a full-stack engineer. Academically, I have a master’s in computer science. Recently, my focus has been on applied AI and agentic workflows.

Along the way I’ve co-founded three companies, including a Google Startups-backed SaaS, contributed to a range of open source projects, presented at JupyterCon three times and co-authored seven peer-reviewed publications.

Recent Posts

All Posts

  1. A hallucinated module, a backfiring RAG pipeline and the MCP server that fixed it

    Sep 13, 2026

    How an eval suite exposed a RAG pipeline that hurt GenePattern Copilot more than it helped, and what finally fixed it.

  2. Orchestrator, planner, six specialists: a multi-agent pipeline that survived production

    Jul 28, 2026

    How a hierarchical multi-agent pipeline increased GenePattern module production fivefold.

Selected Projects

All Projects

GenePattern Module Toolkit

University of California, San Diego  ·  Sep 2025 – Jun 2026

Hierarchical pipeline that wraps bioinformatics tools as production modules. A supervisor coordinates researcher, planner and six artifact agents, linting each artifact and rewinding to whichever agent introduced a failure.

Multi-Agent LoRA Pydantic AI Python

GenePattern Copilot

University of California, San Diego  ·  Apr 2025 – Jun 2026

RAG-based agentic assistant grounded in documentation and help forum. Uses an MCP server to run GenePattern jobs and retrieve results on the user’s behalf. Thumbs-up and thumbs-down ratings feed its evaluation dataset.

RAG Vector Databases MCP Pydantic AI

Agent Template

Open Source  ·  Jul 2026 – Present

Opinionated scaffolding for building AI agents: three interchangeable patterns, Logfire tracing, per-run token caps and an eval suite spanning both pass/fail and LLM-as-judge cases.

Evals Logfire Pydantic AI Python

Content Agent

Open Source  ·  Aug 2026 – Present

Local agent with a web interface that drafts on-brand posts and newsletters for multiple brands, each with its own voice and content history. Inference runs via a self-hosted Ollama model.

Ollama FastAPI HTMX SQLite