Staff Product Manager, Training Infrastructure - Weights & Biases

CoreWeave•San Francisco, NY

9d•$188,000 - $275,000•Onsite

About The Position

Training a frontier model is as much an infrastructure problem as it is a modeling problem. When a run fails at hour 47 of a 72-hour job, the question is almost never "what's wrong with the code" — it's "what happened on the cluster." Node failures, network timeouts, storage bottlenecks, and scheduling decisions determine whether training runs succeed or fail, how much they cost, and how fast teams can iterate. Until now, the developer tools and the infrastructure telemetry have lived in completely separate worlds. The CoreWeave acquisition changes that. W&B can now build products that connect application-level experiment tracking with cluster health, hardware telemetry, and compute orchestration — a surface area that's hard to assemble without owning both the developer tools and the infrastructure. Reporting to the Director of Product Management, this Staff Product Manager will own the strategy and execution for the products at this intersection: W&B Launch (job submission and compute orchestration), infrastructure insights that leverage CoreWeave's unique telemetry and platform capabilities, and new product concepts — including agentic and automated research workflows — that only exist because W&B and CoreWeave are now one company. Sandboxes (managed development environments) are also part of this surface, but the core opportunity is building infrastructure-native developer tools that take advantage of the combined platform. You'll be working directly with frontier model builders — teams training on thousands of GPUs — to understand what they need and ship it. If you're energized by solving infrastructure challenges that directly accelerate deep learning workflows, collaborating with the world’s best AI teams, and delivering tools that move the industry frontier forward, his role is for you.

Requirements

7+ years as a product manager working on developer tools, deep learning infrastructure, or ML engineer platforms.
End-to-end ownership: You've taken a technical product—whether a model, orchestration feature, or developer tool—all the way to production, including defining success criteria, measuring effectiveness, and iterating to improve impact and user happiness.
Technical fluency: You’re comfortable discussing Kubernetes and SLURM, job scheduling, hyperparameter optimization algorithms, and API design with engineers and AI researchers.
Cross-functional influence: You’re skilled at managing and influencing AI engineers, software engineers, and stakeholders—with or without official authority—to align teams and drive outcomes.
Clear communication: A major part of this role is seeking feedback from many directions and channeling it into a single, executable stream. You bring clarity and organization to complex, fast-moving environments.
Care and taste: You have strong opinions about the products you use and a high bar for developer experience.

Nice To Haves

Experience with deep learning frameworks (PyTorch, JAX), distributed training, or LLM development workflows.
Hands-on experience with SLURM or Kubernetes for GPU-intensive workloads.
Prior exposure to W&B or other experiment-tracking and model management platforms.
Understanding of hyperparameter optimization methods and evaluation workflows for large models.

Responsibilities

Define and ship infrastructure-native products that only W&B + CoreWeave can build. This is the core of the role. The integration between W&B's experiment tracking and CoreWeave's infrastructure platform opens a new product surface — giving researchers visibility and control that spans their code, their runs, and the machines those runs execute on. You'll identify the highest-value opportunities in this design space, build conviction around them, and drive the roadmap from concept through launch and iteration.
Own W&B Launch and evolve it for the CoreWeave era. Launch is W&B's job submission and compute orchestration product. Today it helps teams submit training jobs to Kubernetes clusters and cloud providers. As part of a vertically integrated AI platform, Launch can go further — with deeper infrastructure awareness and the ability to serve as the orchestration layer for agentic and automated research workflows. You'll set the vision for what that looks like and ship it.
Spend real time with frontier AI teams and learn how they actually work. Foundation model builders are your primary customers. You'll spend significant time in their workflows — understanding how they manage infrastructure, debug failures, orchestrate computation, and increasingly automate their research processes. These conversations will directly shape the product roadmap.
Build the bridge between W&B product and CoreWeave platform engineering. This role sits at the seam between two organizations that are becoming one. You'll partner with CoreWeave's infrastructure and platform teams to identify integration opportunities, and with W&B's engineering teams to build them into products researchers actually want to use. Your ability to translate between infrastructure engineers and ML researchers will be critical.
Find and resolve friction points in the job submission, compute management, and infrastructure observability workflows, maintaining W&B's high bar for craft and quality.
Explore new product concepts at the hardware-software boundary. The design space opened by the CoreWeave integration is large — spanning profiling, environments, automation, agents, and more. You'll evaluate which new concepts have the highest impact for frontier AI teams and drive the most promising ones from zero to one.