Editor's note: As part of the Bay Area Inno Awards, the San Francisco Business Times and Silicon Valley Business Journal are honoring startups across the region's innovation space. Here's the honoree in the artificial intelligence category.
Weights & Biases may never be a household name like some other tech giants, and that’s OK with CEO Lukas Biewald.
If the company is successful, it will be relatively invisible to most people while it plays a critical role in the evolution of the AI industry.
The San Francisco startup works behind the scenes, developing tools for software developers that make it easier for their code to be tracked, monitored, audited and, by extension, fixed.
Some AI systems are so-called black boxes — opaque repositories of code that might be able to scale quickly but can’t be easily checked or amended even by the team that built it.
Instead, Weights & Biases is making AI models more transparent by making them explainable and observable, which ensures that the data and machine learning models stay organized and accessible even as employees come and go. It also supports compliance with data privacy regulations by making data more easily retrievable.
AI models also require a lot of coding and an immense amount of data and computing power. It makes the costs skyrocket, and their complexity makes it impossible for humans to manually monitor.
So, the solution has to involve automation, as well.
“Keeping track of everything is a hard human problem,” CEO Lukas Biewald said. “We try to make it so easy that the people working on it actually want to do it because it actually helps them with their with their work. And then the organization also improves and safety improves.”
That’s especially vital for flagging errors and correcting for biases, but it can also help organizations improve search results, moderate user generated content and detect defective products as they’re manufactured.
Other use cases include supply chain management, marketing and sales predictions, fraud detection, chat bots and autonomous vehicle development.
The only industries they won’t ever work with are those that are sanctioned or illegal, but otherwise, the platform can be applied to just about any industry, Biewald said.
Weights & Biases even worked with OpenAI in its earlier days, Biewald told me, before it sparked a global generative AI frenzy.
It makes the platform available for free to students and others working on personal projects, and charges for commercial and enterprise-level usage.
“It’s important to me that we make machine learning available to everyone that’s interested in doing it,” Biewald said. “I had the opportunity to go to Stanford, which turned out to be ground zero for this stuff, but that was luck … We have a role to play to help people become productive machine learning engineers.”
Weights & Biases currently has around 200 employees and more than 600,000 users.
The next step? Getting to 1 million users, Biewald said.
“I would love to be the system of record for people building and deploying machine learning models, which maybe means we’re not a household name,” Biewald said, “but I love the idea of being behind the scenes, helping tons of different people use machine learning for lots of different applications.”
About Weights & Biases
Location: San Francisco
Industries: Developer tools, artificial intelligence
Founders: CEO Lukas Biewald, CTO Shawn Lewis and CISO Christopher Van Pelt
Founded: 2017
Funding: $200M
Major investors: Felicis, Insight Partners, Coatue Management, Flight Ventures, Gaingels, Bloomberg Beta, Octave and former GitHub executive Nat Friedman.
Why they were chosen: Weights & Biases is helping software developers add transparency and accountability to AI models so they can be audited, monitored and updated with relative ease.
