Co-founder & CEO |Weights & Biases (CoreWeave)
Serial entrepreneur who built Weights & Biases into the leading MLOps platform for experiment tracking and LLM evaluation, previously founding Figure Eight (sold for $300M).
Biography
Lukas Biewald (born 1981, Boston) is an American entrepreneur and AI tooling pioneer best known as the co-founder and CEO of Weights & Biases (W&B), the leading MLOps platform for experiment tracking, model management, and LLM application development. He earned both a BS and MS in Computer Science from Stanford University, studying under Daphne Koller. After stints at Yahoo (Search Relevance for Yahoo Japan) and Powerset (acquired by Microsoft in 2008), Biewald co-founded CrowdFlower in 2007, a human-in-the-loop data labeling platform later rebranded as Figure Eight and sold to Appen for $300M in 2019. In 2017 he co-founded Weights & Biases with Chris Van Pelt and Shawn Lewis, building experiment tracking tools used by OpenAI, Meta, Samsung, Spotify, and 700+ enterprises. The company raised over $250M across five funding rounds, reaching a $1.25B valuation in 2023. In March 2025, CoreWeave announced its acquisition of W&B for $1.7B (completed May 2025). Biewald also hosts the Gradient Dissent podcast, interviewing leaders in AI and machine learning.
Co-founded the leading MLOps platform for experiment tracking, model versioning, dataset management, and collaborative AI development. Used by 700+ enterprises including OpenAI, Meta, Samsung, and Spotify. The platform lets ML teams log experiments, visualize model performance, and reproduce results with a few lines of code.
Built the foundational experiment tracking product that lets developers log metrics, hyperparameters, code, and system resources during model training. Became the de facto standard for ML experiment management across research labs and enterprises.
Developed an LLM observability and evaluation platform that traces all LLM calls with a single line of code, provides pre-built scorers, and enables systematic iteration on accuracy, latency, cost, and user experience for generative AI applications. Launched GA in December 2024.
Founded one of the earliest human-in-the-loop ML platforms for crowdsourced data labeling. Pioneered the approach of combining human intelligence with machine learning to create high-quality training datasets. Sold to Appen for $300M in 2019.
Hosts a bi-weekly AI podcast featuring in-depth conversations with industry leaders from NVIDIA, Meta, Google, OpenAI, GitHub, and more. Over 128 episodes published since 2020, serving as a key resource for the ML community.
Open-source machine learning lessons and teaching projects designed for engineers, with over 2,500 GitHub stars. Demonstrates Biewald's commitment to ML education and community building.
Software developers have high quality tools for every part of their workflow, but ML practitioners are still in the dark ages. You can't have AI safety if ML teams can't systematically track the models they build and the datasets they use.
The idea that computers could learn to do things on their own just seemed amazing to me.
I really do want to make stuff that people use and like. That sustains me.
I got really excited about building a company that would help behind the scenes in making stuff reliable.
Research generated March 19, 2026