About · ML systems

I build ML systems that turn traces, evaluations, and tools into clearer next actions.

Teaching programming and working in computer-vision research came first. That led into analytics, data science, recommendation systems, and production ML at SAP. The path wasn’t especially tidy, but the same problem kept showing up: making complex technical systems easier to inspect, improve, and use.

At Weights & Biases, part of CoreWeave, I contribute to MCP deployment and reliability, ARIA's streaming and recovery, and evaluations that inspect agent behavior. Lessons from that work become open-source tools, articles, courses, and hands-on workshops.

See the work

How the work developed.

  1. 2021–Now

    Weights & Biases

    ML support → MLOps growth → agent systems

  2. 2017–2021

    SAP & Brilliant Hire by SAP

    Analytics → data science → production ML

  3. 2016–2019

    Temple University

    Teaching & research

For event organizers

Read speaker profile

Anish Shah builds agent infrastructure and evaluation workflows at Weights & Biases, part of CoreWeave. His engineering contributions span the W&B MCP Server's deployment and reliability, ARIA's streaming and recovery, and controlled studies of agent behavior. His talks use that work to explain how to trace failures, test tool interfaces, and decide what the evidence supports.

His path spans programming education, computer-vision research, analytics, recommendation systems, and production ML. He develops hands-on curricula and co-taught Fully Connected's Your Agent Learned to Cheat lab, where participants traced reward shortcuts and repaired the tests that missed them. He brings concrete implementations, experimental limits, and the engineering decisions behind them into each session.

Topics

Browse talks and workshopsContact on LinkedInhttps://www.linkedin.com/in/anish-shah