THE DIRECT ANSWER

Lukas Biewald co-founded Weights & Biases with Chris Van Pelt and Shawn Lewis in 2017. Its company history describes an initial focus on experiment tracking and its move into CoreWeave in May 2025. His profile concerns a less visible part of AI progress: preserving the evidence behind a model or application improvement.

Person
Lukas Biewald
Company
Weights & Biases
Area of work
Experiment tracking and reproducible evaluation

From experiments to infrastructure

Weights & Biases says Lukas Biewald, Chris Van Pelt and Shawn Lewis founded the company in 2017. Its early focus was experiment tracking. The company became part of CoreWeave in May 2025, and its current leadership page lists Biewald as general manager and co-founder.

Public record: Weights & Biases: founder biography or company record

Why experiment records matter

Two model runs can differ in data, parameters, code or environment. If those differences are not recorded, an apparent improvement can be difficult to explain or reproduce. Experiment tracking helps connect a result with the conditions that produced it. Weights & Biases' history places this problem at the start of the company. For a reader following Biewald, it is a reminder that AI development includes disciplined record-keeping as well as algorithm design. A chart without the underlying configuration can be an incomplete piece of evidence.

How the problem extends to AI applications

An application using a hosted model still changes over time. Prompts, retrieved documents, tool implementations and provider versions can all affect behaviour. A team needs to know which combination was evaluated and which one is currently running. Our analysis is that application evaluation inherits many of the same reproducibility problems as model training. A useful record links representative inputs, expected behaviour and observed failures to a versioned system, making a later regression easier to investigate.

A simple evaluation ledger

For each release, record the task definition, dataset revision, model identifier, relevant settings and acceptance criteria. Keep failed examples alongside aggregate scores so an average does not hide a serious error. Separate development examples from a held-out set used for the final comparison. When a result improves, identify what changed and what became worse. This is a proposed workflow, not a report of using Weights & Biases. It explains why observability infrastructure can matter even when it is less visible than the application people interact with.

COMMON QUESTIONS

Questions about Lukas Biewald

Is Weights & Biases still an independent company?

Its company page states that it became part of CoreWeave in May 2025. The same page identifies Biewald as general manager and co-founder.

What should an AI experiment record include?

At minimum, preserve the system configuration, data version, evaluation method and observed results. The exact fields depend on what can change the behaviour being measured.

SOURCES & NOTES

Follow the evidence.

This AI-assisted profile combines the linked public record with Promptwire's editorial analysis. It is not an interview or a hands-on product review. Suggested tests are proposals, not reported results. Dated roles describe the period identified in the text. Company claims are attributed to their source.

  1. Weights & Biases: founder biography or company record

Public sources support the factual background. Analysis and illustrative exercises are identified in the text. How we assess AI · Editorial standards

Spot a factual error or an outdated detail?