AI development tooling startup Weights & Biases reels in $50M

AI development tooling startup Weights & Biases reels in $50M

Posted on



Weights & Biases Inc., a startup helping Nvidia Corp. and OpenAI LP build artificial intelligence models faster, today announced that it has raised $50 million in funding.

Former GitHub Chief Executive Officer Nat Nat Friedman and former Y Combinator partner Daniel Gross led the round. A half dozen institutional investors, most of them existing backers, contributed as well. Weights & Biases is now valued at $1.25 billion, or $250 million more than after its previous funding round in late 2021.

Since that funding round, the startup’s installed base has ballooned from 100,000 users to 700,000. Weights & Biases Says that its software has tracked about 300 million hours’ worth of AI experiments for customers. The startup lists OpenAI, Nvidia, Microsoft Corp. and Meta Platforms Inc. among its customers along with numerous AI startups. 

Machine learning projects have a strong trial and error element to them. During development, engineers usually create many replicas of a neural network and make slight modifications to each version. They then compare the different versions to find the one that is best aligned with project requirements.

Such tests generate a significant amount of technical data. As a result, manually analyzing that data to compare how different versions of a neural network perform can be prohibitively time-consuming. Weights & Biases offers a platform that can collect the results of AI tests and visualize them to help engineers more quickly compare different model versions.

The neural network testing phase accounts for a significant portion of the typical AI project’s duration. As a result, speeding up the process allows companies to build machine learning applications faster. Weights & Biases says that its platform helped OpenAI, its first customer, speed up GPT-4’s training. 

The startup has rolled out several additional tools over the past few years. One tool makes it easier for developers to deploy new neural networks on their company’s production infrastructure. Another automates hyperparameter optimization, or the process of fine-tuning an AI model’s settings in a way that increases its speed and accuracy.

“Our goal at Weights & Biases has never changed: to build the best tools for machine learning practitioners,” said Weights & Biases co-founder and Chief Executive Officer Lukas Biewald. 

Weights & Biases will use its latest funding round to enhance its product portfolio. That effort will encompass, among other offerings, a new tool called Prompts the startup announced in conjunction with today’s funding round.

Prompts is designed to help companies troubleshoot issues in their large language models. According to Weights & Biases, it works by logging each prompt that users enter into a model and the output the AI generates in response. The tool automatically flags user inputs that lead to an error.

In some AI applications, prompts are processed by not only the built-in neural network but also various other software components. According to Weights & Biases, its Prompts tool highlights which components each input goes through while it’s processed and in what order. Developers can use that data to inform their troubleshooting efforts.

Image: Weights & Biases

Your vote of support is important to us and it helps us keep the content FREE.

One-click below supports our mission to provide free, deep and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *