Weights & Biases
Weights and Biases builds cutting-edge tools for machine learning. Our software makes it easier to build production-ready machine learning models at scale. We help ML teams visualize performance and iterate faster.
We have extensive experience working with companies to turn machine learning research projects into scalable, real-world deployments. We’ve watched hundreds of teams struggle to deploy machine learning models successfully, and the same problems show up repeatedly. Machine learning has created a fundamentally new kind of programming, requiring a fundamentally new set of developer tools. We created Weights & Biases to provide that missing toolkit.
Two of our founders, Lukas Biewald and Chris Van Pelt, previously founded CrowdFlower (now Figure Eight), a successful startup that does ML data labeling, with mid-eight-figure ARR. We're using the connections we built via CrowdFlower to make sure we're working with the best potential customers from the start.
We're working closely with OpenAI, Toyota Research Institute, and others to develop a product that helps their teams build better models.
Here's a quote from Wojciech Zaremba, Cofounder and Robotics Lead, OpenAI:
W&B allows to scale up insights from a single researcher to the entire team, and from a single machine to hundreds of them.
We value autonomy and want team members to feel a sense of ownership over product and engineering direction. We love when someone jumps in and cleans up something that's broken, even if it's not their fault. We value automated testing so that we can be comfortable making large changes. We love good type systems. We value shipping and incremental improvement.
We work together to come up with weekly goals for each of us, then we execute them. For longer projects we come up with a plan and then check in every week during a product meeting. We have a weekly company meeting where we talk through our quarterly goals, sales, and everything else.
Many of our challenges are product design related, whether it's building data exploration and visualization features for thousands of experiments in the frontend, or designing clean APIs for machine learning practitioners.
We handle a respectable amount of data — our customers run many distributed training experiments in parallel, with hundreds of machines reporting event information per experiment. This will of course continue to scale as we grow.
Build a UI to allow a user to interactively guide hyper-parameter sweeps.
Swap out the datastore used for user log data so we can scale.
Build a new tool for managing the evolution of large datasets.
We love coming to work just to be around smart, driven people. We value hard work while having fun at the same time. We love creativity of thought and spirit. We believe AI is philosophically interesting.
We have an unlimited PTO policy. Everyone gets their jobs done and has trust in each others ability to schedule their time appropriately.
About once a month we rent an Airbnb somewhere as a team. We hangout, cook and build stuff together. This will never be mandatory but has been a great way to bond and make rapid product progress without distractions.
Employees are encouraged to go to workshops and conferences that further their development.
Top tier PPO health insurance.
We have a lot of fun together as a team.
Interested in this company?
Skip straight to final-round interviews by applying through Triplebyte.