Robb Doering

Robb Doering

Senior Full Stack Engineer · Agentic Engineering

robb@doering.ai+1 415 713 1766doering.ailinkedin.com/in/robbdgitlab.com/robbdgithub.com/robbwdoering

Download my résumé

Summary

Senior full stack engineer (TypeScript, React, Python) with 6+ years of professional experience, from Google AdSense to satellite command-and-control software. Three years of independent AI work became full-time agentic engineering: directing coding agents in isolated Git worktrees and building the MCP servers, code review, guardrails, and recovery tooling around them. I relish any opportunity to apply science or philosophy to engineering challenges.

Experience

  1. Aug 2023 – Sep 2026

    Doering Thinking Machines Software Engineer (self-employed)

    AI prototyping and agent tooling in Python and TypeScript, published as ~30 open-source repositories.

    • Agent orchestration. Directed coding agents (Claude Code, Codex CLI, Gemini CLI) through scoped tasks in isolated Git worktrees, with automated code review as the gate to main: nearly 2,400 agent branches merged across 45 repositories in Q3 2026.
    • MCP integrations. Built four FastMCP servers that replace fragile shell calls with structured tools: inTact (on PyPI) heals stale agent edits through seven increasingly permissive matches and refuses ambiguous ones; the others handle Git with commit-time secret scanning, Plane task tracking, and task routing.
    • Code review and guardrails. Built myGate, which reviews each GitLab merge request from an immutable Git snapshot, posts findings as discussions, and holds the merge until major findings are resolved: 1,500+ merges reviewed from mid-August through September 2026. Its hooks block costly nested agent launches.
    • Recovery. Built tooling to resume interrupted agent runs: session analytics classify how each run ended (frozen on a tool call, aborted), and a resumer agent rebuilds the work from transcripts, task records, and repository state, then finishes only what remains.
    • Engineering standards. Codified the workflow as 29 policy documents with citable rule IDs and 46 agent skills, which myFleet projects from one canonical copy into eleven agent harnesses. Held the libraries to GitLab CI, deep type coverage, and Sphinx documentation, with ~9,900 pytest cases across four of them.
    • Models and retrieval. Ran a self-hosted LiteLLM gateway with sealed credentials; built embedding search on PostgreSQL; served models with vLLM and llama.cpp; fine-tuned with QLoRA on an RTX 4090.

  2. Feb 2022 – Jun 2023

    Google Software Engineer, AdSense

    Modernized the JavaScript ad tag of AdSense, a multi-billion-dollar-a-year product, and implemented new ad formats across browser (TypeScript) and backend (C++) code.

    • Owned features from start to finish at Google scale, meeting strict performance requirements and shipping through Google's production deployment systems.
    • Set up, monitored, and analyzed A/B experiments on billions of ad queries to verify the impact of my changes, using Python notebooks and internal tools.
    • Investigated and documented JavaScript behavior for scripts that run in over 2M browser environments.

  3. Jun 2019 – Jul 2021

    Space Ground System Solutions Software Engineer

    Helped bring a command-and-control system for government satellites to the web, building features in React/Redux, Java, and C++ on data-intensive APIs from NASA, DoD, and commercial partners.

    • Delivered ~12 full stack features end to end, some from design documents and others where I led requirements gathering and UX design.
    • Chosen as acting project lead for six engineers during the lead's three-month leave; ran Agile ceremonies, met with stakeholders, and cleared blockers.
    • Introduced Jest unit tests to a lightly tested codebase and helped coordinate regular functional testing.

  4. May 2018 – Aug 2018

    Fujitsu Network Communications Automation Tools Intern

    Built a Java/Spring MVC and JavaScript web application that formalized a project-management procedure and collected its data for executive analysis.

Projects

  1. Astro · React · OCR

    The Epstein Index Public Document Explorer

    • A public explorer for over 45,000 DOJ case-file documents, fed by an eight-stage Python pipeline (download, OCR, clean), with an interactive React timeline, maps, and a network graph.

  2. Python data pipeline

    Wikiparse Wikitext Translation Pipeline

    • Extracted 45M citations from a full English Wikipedia dump on home hardware and published a Hugging Face dataset of 5.1M articles; converts wikitext to Markdown with composed regular expressions.

  3. Astro · TypeScript

    doering.ai Professional Portfolio Site

    • An Astro and TypeScript site on Bun and Tailwind CSS, with accessibility and reproducibility checks.

  4. Python library

    myBasis Ergonomic Python Utilities

    • A broad extension of the Python standard library, published on PyPI as my-basis, centered on runtime type coercion, text processing, and functional programming.

  5. Python · Rust · LSP

    Sublime Text tooling Plugins, Language Servers, and Test Infrastructure

    • A mock editor API for headless pytest, a Rust-backed language server for .sublime-syntax grammars, and LSP-pyrefly for Meta's pyrefly type checker.

Skills

Languages
Python, TypeScript/JavaScript, C++, Java, SQL, Bash
Frontend
React, Redux, Next.js, Astro, HTML/CSS, Tailwind CSS, Bun
Backend and data
REST APIs, WebSockets, Node.js, Flask, PostgreSQL, embeddings, RAG
Agentic engineering
Claude Code, Codex CLI, Gemini CLI, MCP servers, skills, hooks, Git worktrees
Models
LiteLLM, vLLM, llama.cpp, Hugging Face Transformers, QLoRA fine-tuning, Marker OCR
Testing and operations
pytest, Jest, A/B testing, GitLab CI/CD, Linux, Docker/Podman, systemd services

Education

  1. Aug 2015 – May 2019

    Vanderbilt University B.S., Computer Science

    Nashville, TN

    Minors in Philosophy and Spanish.

  2. Sep 2020 – Jun 2022

    Georgia Institute of Technology M.S. coursework, Computer Science · Interactive Intelligence

    Online

    • Completed eight of ten courses; degree remains incomplete.
    • Coursework: Networks, Operating Systems, Educational Technology, Human-Computer Interaction, Artificial Intelligence, Knowledge-Based AI, Software Development Processes, and Visual Analytics.

Working with me

I'm available for full-time roles first, and for contract or fractional engineering where the problem is interesting: fullstack product work in Python and TypeScript, developer tooling, agent and model-inference systems, and the documentation, specification, and validation work that usually goes undone. I work remotely from Atlanta, GA, on US hours. Email is the fastest route.

How I use AI

I use AI models every day, as tooling, and I think you should know how. Every claim on this site derives from repository data or my own writing; code changes on my projects go through merge-request review, tests, and CI whether a person or an agent drafted them. My cover letters are built from sentences I wrote, and a model helps me assemble and clean them; I disclose model assistance whenever I'm asked. I'm glad to talk about where these tools belong in your team's work and where they don't.

Questions

What is Doering Thinking Machines?
The name I worked under from August 2023 to September 2026: self-directed engineering and research, small contracts, and the personal software fleet catalogued on this site — 34 repositories, 34 with public source.
Remote only?
Yes — from Atlanta, GA, on US hours. A hybrid arrangement is worth a conversation; relocation is not on the table right now.
Which stack?
Python and TypeScript first — pydantic, pytest, and uv on one side; React, Astro, and Node on the other — with C++ and Java in the record, PostgreSQL, Docker and Podman, and the model-inference stack around vLLM, DSPy, and MCP.
Do you take contract work?
Yes. See Working with me above.