Tim Sweeney

Principal Software Engineer · Architect for AI Agents, LLM Observability, and ML Infrastructure

Principal engineer who repeatedly takes AI products from first prototype to production: W&B Model Registry, one of the first LLM trace viewers, Weave, and CoreWeave ARIA. Ten years across ML infrastructure, product and engineering, and a committed teacher of how these systems work. M.S. Computer Science (Machine Learning, focus on reinforcement learning), Georgia Tech.

Expertise: LLMs and agentic systems · AI evaluation and observability · platform architecture · developer platforms and API design · ML infrastructure · reinforcement learning · Python and TypeScript

timssweeney.com GitHub LinkedIn

Experience

  1. 2025 — now

    CoreWeave

    Principal Software Engineer · Tech lead, CoreWeave ARIA

    • Led CoreWeave ARIA, an AI research agent for ML teams, from first prototype to general availability in CoreWeave Forge in eight months: 10,000+ users and over 99% successful agent turns.
    • Architected the agent runtime, sandboxed execution, LLM gateway, billing, spend controls and security; owns reliability and on-call.
    • Founded the cross-team agent working group; made ARIA a platform other Forge products embed, running on CoreWeave's own inference and sandboxes. Pioneered agent-driven engineering: ARIA is built, tested and monitored largely by AI coding agents.
  2. 2020 — 2025

    Weights & Biases acquired by CoreWeave, 2025

    Principal Software Engineer, 2023 — 2025 · Software Engineer, 2020 — 2023

    • Co-led Weave from prototype to a revenue-generating LLM evaluation and observability product; set its architecture, roadmap and technical direction and scaled its six-person engineering team.
    • Built most of Weave's evaluation product line (scorers and guardrails, leaderboards, comparisons, playground) as weekly returning users grew about 3× in 2025. Open source: 1,000+ GitHub stars, 790+ dependents.
    • Built one of the first LLM trace viewers in 2023, weeks after GPT-4's release, the starting point for W&B's tracing products. Earlier conceived and shipped W&B Model Registry from idea to launch.
    • Partnered with sales to close several of Weave's largest enterprise deals; forward-deployed with the Aston Martin F1 partnership.
  3. 2018 — 2020

    Twitter

    Product Manager II, Machine Learning Platform

    • Owned strategy for Twitter's internal ML platform; drove 92% adoption of its TensorFlow-based core, contributing to a 1.7% revenue increase in 2018 and 350% faster model iteration in 2019.
  4. 2015 — 2018

    Workday

    Product Manager, Enterprise Search and ML · promoted twice in under two years

Teaching & mentorship

  1. 2020 — now

    Taught a monthly W&B 101 course; leads architecture deep dives, live debugging sessions and sales enablement; mentored an engineer into tech lead and coached a product manager; conference speaker (AI Engineer World's Fair, W&B Fully Connected).

Talks & publications

  1. 2026

    An AI research agent that runs your experiments AI Engineer World's Fair timssweeney.com/talks/aria

  2. 2019

    Ranking Tweets with TensorFlow TensorFlow Blog timssweeney.com/writing/ranking-tweets

  3. 2018

    Embeddings at Twitter Twitter Engineering Blog timssweeney.com/writing/embeddings

Education

  1. 2018 — 2020

    Georgia Institute of Technology

    M.S. Computer Science · Machine Learning specialization, focus on reinforcement learning · GPA 4.0

    Coursework: Reinforcement Learning, Machine Learning, Artificial Intelligence, AI for Robotics, Computer Vision, Graduate Algorithms, Big Data, Time Series Analysis

  2. 2011 — 2015

    Cal Poly, San Luis Obispo

    B.S. Business Administration, Information Systems; minor in Computer Science · GPA 3.70 · President, Cal Poly Entrepreneurs

Last updated October 2026