Making the most of AI? You could be.

I help technology leaders and their teams put AI where it pays off. I've spent 2 decades building software and leading engineers, and I still build.

Lead, Staff, and Principal engineering roles for the past decade.

Problems I help with

  1. “Our AI spend isn't moving the needle.”

    Where you get stuck

    • We generate more code but we don't ship more.
    • Our processes are still the bottleneck.
    • Our engineering velocity has not improved reliably with AI.

    How I help

    I find where AI can take real work off your teams, then change the process and tooling to match.

  2. “Everyone has an AI pitch, and we don't know which to choose.”

    Where you get stuck

    • Lots of pilots, nothing that changed the business.
    • Vendors setting the agenda.
    • Build or buy paralysis.

    How I help

    I sort the options into what pays off now, what's worth a small test, and what to skip, then help you decide whether to build, buy, or wait.

  3. “The board wants our AI strategy, and we don't have a clear one.”

    Where you get stuck

    • The strategy is a list of tools.
    • Every team doing its own thing.
    • AI in our product feels bolted on for its own sake.

    How I help

    A plan you can defend: where AI goes, what it costs, and how you'll know it's working.

  4. “Some of our people barely use AI, and we don't know what to do about it.”

    Where you get stuck

    • A few power users, a lot of holdouts.
    • Engineers view AI usage as all or nothing.
    • Engineers view AI as a threat, not an asset.

    How I help

    I find out why people hold back, fix those causes, and leave behind practices your new hires pick up too.

  5. “Our AI features work in the demo and break in production.”

    Where you get stuck

    • Answers that change from run to run.
    • No way to tell if a model update made things worse.
    • LLMs used where plain code would do better.

    How I help

    I build the structure around the model: deterministic logic where it has to be right, and evaluations that catch regressions before your customers do.

Track record

  • Architecture

    Led architecture for Upwork and its 240-engineer, $175M Marketplace business unit.

  • AI evaluation

    Designed and built a model and prompt evaluation service used by ML and central engineering teams.

  • Cloud migration

    Proposed and led an on-prem to cloud migration: 3 monoliths, 37 services, 600 jobs, and streaming.

  • Cost

    Cut $750k+ a year from infrastructure spend across a 550-engineer organization.

  • New product

    Created the system design for a newly launched product and owned its security and compliance.

  • Teaching

    Built and taught engineering courses in system design, Python, and Kafka, and lectured at PyCon EduSummit.

Core skills

  • AI adoption strategy and team buy-in
  • Model and vendor evaluation
  • AI-powered automations
  • Getting the most from Claude Code and Codex
  • Custom harnesses and agentic loops
  • Weaving deterministic logic into LLM services and data flows

Industry expertise

Fintech and regulated environments

  • Architect for the finance technology behind $14B in revenue: 9 teams, 110 engineers.
  • SOX-compliant Kafka and company wide message queue technology strategy and migration plan.
  • Owned SOC 2, ISO 27001 and 27018, and SOX compliance for a large company pivot 0-1 product.

Marketplaces and e-commerce

  • Redesigned pricing engine & services to process 60M+ option combinations across 20 million products.
  • Set technical strategy and OKRs for a two-sided marketplace leading over 200 engineers.
  • Career long history of close partnership with ML platform and engineering teams.

Also: advertising and media platforms, and healthcare data.

How we work together

Assessment

A focused look at how your teams work with AI. You get help prioritizing what to try, implement, buy, or skip.

Advisory

An ongoing seat at the table for strategy, architecture review, and build or buy calls, tied to what the business needs.

Hands-on

I work alongside your engineers on the hard parts: harnesses, agentic loops, evaluations, and the deterministic logic around the model.

Training

Workshops and pairing that raise how your engineers use AI across the development cycle, without giving up quality or safety.

I regularly push the limits of what is feasible with AI. Some of the results are open source: Gossamer (a programming language), LiteAnvil (a code editor), Waftty (a terminal emulator). Some are not (a custom harness).