skip to content
Steel Ferguson

About

Steel Ferguson

I build AI agents and the evaluation systems that keep them trustworthy at scale.

I work on AI agents and evaluation at Shopify, where I'm a core contributor to the Shop.com conversational agent, designed the offline eval framework that gates rollouts, and recently shipped a distillation that's projected to save more than $4M annually. Before that I spent three years on Meta's Facebook Integrity team building ML systems at billions-of-users scale to detect and mitigate coordinated abuse.

On the side I'm the co-founder and only engineer of Basin Climbing, a real climbing gym in Waco, TX serving roughly 14,000 customers. I built and run the entire platform: the iOS member app, two AI agents (customer-facing and analytics), a CRM, an agentic email-flow generator, the data pipeline, and the dashboards.

I care most about AI safety, evaluation methodology, and the production end of the agentic stack. The integrity work at Meta and the agent eval work at Shopify share a throughline: building the systems that catch failure modes before they reach users.

Experience

  1. Senior Applied Machine Learning Engineer

    Jul 2025 – Present

    Shopify — Remote

    Core contributor to the consumer-facing AI agent on Shop.com. Designed the offline eval framework (LLM-as-judge, benchmark suites, agent observability) that gates rollouts. Built a two-step agentic conversation-starter system, custom judge pipeline, and a Qwen-4B distillation projected to save >$4M annually. Drove DSPy/GEPA programmatic prompt optimization. Architected buyer personalization for 30M+ users.

  2. Co-Founder and Chief Technology Officer

    Jun 2024 – Present

    Basin Climbing and Fitness — Waco, TX

    Sole engineer. Shipped a React Native iOS app (live on the App Store), two production AI agents on a multi-agent / MCP architecture with end-to-end eval and observability, a CRM with two-way SMS and a follow-up engine, a Klaviyo flow generator/manager, and the data infrastructure (multi-source pipeline + executive dashboards). Partner with the CEO on company strategy.

  3. Senior Machine Learning Engineer, Facebook Integrity

    Jul 2022 – Apr 2025

    Meta — Remote

    Shipped a deep learning classifier detecting coordinated automated behavior (scripted friending), reducing it by 10%. Built impersonator detection paired with a deployed mitigation that hid friend lists from risky accounts. Restructured the integrity-filtering layer of the recommender system in a change that lifted Facebook MAU by 10M. Production ML systems serving billions of users across recommendation and integrity surfaces.

Earlier

  1. Team Lead and Senior Data Scientist, Progressive Leasing

    2021 – 2022

    Led a team of three. Built fraud detection models and pipelines that delivered $2.5M in recovered profits across 5M customers annually.

  2. Instructor of Data Science, UC Berkeley (remote)

    2021 – 2022

    Taught ML and statistical modeling to 80+ working professionals; managed six teaching assistants.

  3. Senior Business Analyst, Capital One — Plano, TX

    2015 – 2021

    Built ML-driven pricing optimization for vehicle pricing ($5M incremental net proceeds). Designed and analyzed 20+ A/B experiments on production user-facing systems.

Education

M.S. Computer Science, Machine Learning specialization

Georgia Institute of Technology · 2022

B.S. Economics

Brigham Young University · 2014

Get in touch

Email steelferguson@gmail.com or DM on LinkedIn.