Boston, you do not want to miss this! The first time we shipped an internal agent swarm, it broke in ways we didn’t even have language for yet. Agents calling agents in loops. Tools silently timing out. Traces that made zero sense until you stared at them for an hour. We figured it out. Then we helped customers figure it out too. Now the wandb team is bringing those lessons to Boston Tech Week. Join Emmanuel Turlay, Uma Krishnaswamy, and Nicolas Remerscheid for a deep dive into what actually works when building multi-agent systems with W&B Weave + MCP: • Real production patterns • Common failure modes • Debugging strategies • Lessons learned the hard way You know… the stuff nobody puts in those launch posts. 📍 100 High Street 🕐 1PM ET Food, drinks, and a room full of engineers building this stuff for real. If you’ve ever stared at a trace wondering “what just happened?” you should be in this room. RSVP here: https://lnkd.in/geJqfMKB #BosTechWeek #AIEngineering #Agents #MCP #LLMOps
Weights & Biases
Software Development
San Francisco, California 92,269 followers
The AI developer platform.
About us
Weights & Biases: the AI developer platform. Build better models faster, fine-tune LLMs, develop GenAI applications with confidence, all in one system of record developers are excited to use. W&B Models is the MLOps solution used by foundation model builders and enterprises who are training, fine-tuning, and deploying models into production. W&B Weave is the LLMOps solution for software developers who want a lightweight but powerful toolset to help them track and evaluate LLM applications. Weights & Biases is trusted by over a 1,000 companies to productionize AI at scale including teams at OpenAI, Meta, NVIDIA, Cohere, Toyota, Square, Salesforce, and Microsoft. Sign up for a 30-day free trial today at http://wandb.me/trial.
- Website
-
https://wandb.ai/site
External link for Weights & Biases
- Industry
- Software Development
- Company size
- 201-500 employees
- Headquarters
- San Francisco, California
- Type
- Privately Held
- Founded
- 2017
- Specialties
- deep learning, developer tools, machine learning, MLOps, GenAI, LLMOps, large language models, llms, Generative AI, Developer Tools, Experiment Tracking, AI Governance, Model Monitoring, Inference, Open Source AI, Model Comparison, Evals & Scorers, Data Quality, Generative AI, AI Observability, Agentic Workflows, RAG (Retrieval-Augmented Generation), Prompt Engineering, Hyperparameter Tuning, Benchmarking, Large Language Models (LLMs), Reproducibility, Dataset Versioning, and Tracing
Products
![Click here to view Weights & Biases]()
Weights & Biases
Machine Learning Software
Weights & Biases helps AI developers build better models faster. Quickly track experiments, version and iterate on datasets, evaluate model performance, reproduce models, and manage your ML workflows end-to-end.
Locations
-
Primary
Get directions
400 Alabama St
San Francisco, California 94110, US
Employees at Weights & Biases
Updates
-
Weights & Biases reposted this
When I told my 2 year old that we have many sandboxes at work now, he couldn’t believe me. Then, he rejoiced, when I told him that I now have a safe playground 🛝 and don’t have to worry about accidents or malicious behavior.
🚀 CoreWeave Sandboxes is now in public preview! It is an execution layer for RL, agent tool use, and model evaluation at scale. When an AI agent runs `rm -rf /` during training, a command that recursively force deletes every file from the root directory of a Linux system, the sandbox contains it. The agent learns what the command does. The training run keeps going. Serverless is the fastest way in. pip install wandb[sandbox], authenticate with your existing W&B key, and you are running sandboxes in minutes. Every CPU sandbox runs inside a Kata Container by default, with its own hardware virtualized kernel, filesystem, and network. No clusters to provision. No infrastructure decisions to make. This matters because models do not just predict text anymore. They write code and run it. They call tools. They navigate browsers. Reinforcement learning loops spin up thousands of sandboxes in parallel per training step, each one executing whatever the agent decides to try. Containment cannot be optional. Because it runs through W&B, sandbox lifecycle events land in the same run timeline as your metrics, and Weave traces connect every model call and tool call back to the sandbox that produced it. Teams running their own CoreWeave Kubernetes clusters can use sandboxes there too. Get started with pip install wandb[sandbox], more info in the comments!
-
Weights & Biases reposted this
Speaking at STAC Research Summit next week. I’ll be sharing how we at Weights & Biases by CoreWeave are working with quant teams building agentic AI systems for large-scale research and what it takes to make these systems reliable in real-world environments. Looking forward to connecting with the #quant community to discuss scale, evals, observability and reliability. Come say hi if you are attending 👋 Chad Travis Dan Deutsch Alan Zaccone Sasha Manuilova Katherin Madche #STACSummit #QuantFinance #AgenticAI
-
-
🚀 CoreWeave Sandboxes is now in public preview! It is an execution layer for RL, agent tool use, and model evaluation at scale. When an AI agent runs `rm -rf /` during training, a command that recursively force deletes every file from the root directory of a Linux system, the sandbox contains it. The agent learns what the command does. The training run keeps going. Serverless is the fastest way in. pip install wandb[sandbox], authenticate with your existing W&B key, and you are running sandboxes in minutes. Every CPU sandbox runs inside a Kata Container by default, with its own hardware virtualized kernel, filesystem, and network. No clusters to provision. No infrastructure decisions to make. This matters because models do not just predict text anymore. They write code and run it. They call tools. They navigate browsers. Reinforcement learning loops spin up thousands of sandboxes in parallel per training step, each one executing whatever the agent decides to try. Containment cannot be optional. Because it runs through W&B, sandbox lifecycle events land in the same run timeline as your metrics, and Weave traces connect every model call and tool call back to the sandbox that produced it. Teams running their own CoreWeave Kubernetes clusters can use sandboxes there too. Get started with pip install wandb[sandbox], more info in the comments!
-
Kimi K2.6 is on CoreWeave Serverless Inference today along with many other top open models, served on CoreWeave’s NVIDIA GB200s. Everything is OpenAI-compatible API, wandb/ model prefix. Works with LiteLLM out of the box.
CoreWeave is now the leading inference provider for Kimi K2.6, independently benchmarked by Artificial Analysis across 11 providers. Kimi K2.6 is one of the most widely used open-weight models globally — favored for coding agents and multi-step agentic frameworks. Every step is an inference call. The workload demands speed, scale, and reliability. There's no room for tail latency, throughput bottlenecks, or unpredictable costs. Open weights are useful. Open weights paired with a serving stack engineered for the workload are what make production economics work. For teams running K2.6 in production, the tradeoff between speed and cost is over. See for yourself. https://lnkd.in/gJCTpGtD
-
-
Weights & Biases reposted this
What could be better than marimo at NY Tech Week? marimo at NY Tech Week...with ice cream! Come join us for a chill (😉) hangout, where you'll get to meet members of our team and community. Hear from the creator of marimo-pair while enjoying a sweet treat. We'll see you in NY! Link in comments to register ⬇️ #NYTechWeek #python #data #developer #openSource
-
-
Weights & Biases reposted this
Europe doesn’t need more people saying we should build AI companies. It needs more people helping the builders actually do it. We just got back from a live Founders and Builders podcast episode at Project Europe’s Barcelona hackathon, with 90 of Europe’s best engineers/builders hacking through the night. CoreWeave were there too, after signing Anthropic and Jane Street as clients that same week 👀 So I wanted to sit down with Ben Richardson, VP Strategy, to understand why the fastest growing AI Cloud in the world had flown a team out to Barca to work with these founders, helping them think, build, and access the infrastructure they needed. We got into: • Why AI needed a purpose-built cloud <- this was fascinating and not obvious to me • What founders should understand about infrastructure as they scale • CoreWeave's approach to M&A and what most founders usually miss • How you can get involved in the next AI hackathon My favourite idea from the conversation was that the value of a hackathon isn’t just the projects built overnight. It’s the co-founder conversations, the relationships, the energy, and the ideas that might become companies six months from now. That’s what Project Europe and CoreWeave are creating. Thanks again to all the partners that came down for the hack: THEKER Robotics, Clay, Preply, Biorce, Cala, 20VC and we're super happy to have CoreWeave as a partner for Founders & Builders.
-
Weights & Biases reposted this
Training robotics AI is complex by default because every modality adds another layer of uncertainty. These systems operate in the physical world, generating rich, multimodal outputs like videos, trajectories, segmentation maps, and simulation results across countless experiments. Comparing performance has traditionally been slow, manual, and error prone. As workflows scale, it becomes easier to miss subtle regressions or overlook key insights, slowing progress when speed matters most. To move faster, robotics teams need better ways to evaluate and compare experiments, with advanced visualization and a unified platform to track and analyze model behavior without switching between tools. Weights & Biases brings AI and high fidelity simulation into a single development view, enabling faster iteration, deeper insights, and more rigorous evaluation for physical AI teams. https://loom.ly/f5c_0ro #icra #robotics #WeightsAndBiases
-
-
Weights & Biases reposted this
Tomorrow's the day! I’ll be at #AWSSummit #London with the Weights & Biases team on Booth B16. Come find us for a live demo, a conversation about your #AI stack, or just to say hello (and pick up one of these cuties 🐝 ) If you're working on #ML or #GenAI on Amazon Web Services (AWS), we'd love to meet you. Book a meeting here: https://lnkd.in/eAfBAbxk
-