As the tech world moves closer to the idea of artificial general intelligence, we’re seeing another interesting theme emerge in the ongoing democratization of AI: a wave of startups building tech to make tech smarter. more accessible to a wider range of users and organizations.
Today, one of them, Baseten – which develops technology to facilitate the integration of machine learning into a company’s operations, production and processes without the need for specialized engineering knowledge – announces $20 million in funding and the official launch of its tools.
These include a client API and library of pre-trained models to deploy models built in TensorFlow, PyTorch, or scikit-learn; the ability to create APIs to power your own applications; and the ability to create custom user interfaces for your applications based on drag-and-drop components.
The company has been running in a closed, private beta for about a year and has so far built up an interesting group of customers, including Stanford and the University of Sydney, Cockroach Labs, and Patreon, among others, who use it for, for example, helping organizations automatically detect abuse (through content moderation) and prevent fraud.
The $20 million is being discussed publicly for the first time now to coincide with the commercial launch, and it’s in two installments, with equally notable names among those backers.
The seed was co-led by Greylock and South Park Commons Fund, with participation also from the AI Fund, Caffeinated Capital and individuals such as Greg Brockman, co-founder and CTO of general intelligence startup OpenAI; Dylan Field, co-founder and CEO of Figma; Mustafa Suleyman, co-founder of DeepMind; and DJ Patil, former Chief Scientist of the United States.
Greylock also led Series A, with participation from South Park Commons, Stripe’s first executive, Lachy Groom; Dev Ittycheria, CEO of MongoDB; Jay Simon, ex-president of Atlassian, now at Bond; Jean-Denis Greze, CTO of Plaid; and Cristina Cordova, another former Stripe executive.
Tuhin Srivastava, co-founder and CEO of Baseten, said in an interview that the funding will be used in part to recruit more technicians and products, and to accelerate its marketing and business development.
The problem that Baseten has identified and is trying to solve is central in the evolution of AI: machine learning tools are becoming more ubiquitous and used, thanks to cheaper computing power, better access to models training and a growing understanding of how and where they can be used. But one area where developers still have to make a big leap forward, and where companies still have to make big investments, is when it comes to actually adopting and integrating machine learning: there remains a large range of technical knowledge that developers and data scientists need to really integrate machine learning into their work.
“We were born from the idea that machine learning will have a massive impact on the world, but it is still difficult to extract value from machine learning models,” Srivastava said. Difficult, because developers and data scientists must have specific knowledge on how to manage machine learning operations, as well as technical expertise to manage upstream and downstream production, he said. “This is one of the reasons why machine learning programs in companies often have very little success: it takes too much effort to get them into production.”
It’s something that Srivastava and his co-founders Amir Haghighat (CTO) and Philip Howes (Chief Scientist) experienced firsthand when they worked together at Gumroad. Haghighat, who was head of engineering, and Srivastava and Howes, who were data scientists, wanted to use machine learning within the payments company to help with fraud detection and content moderation , and realized that they needed to retrieve a lot of additional information. stack the engineering skills – or hire specialists – to build and integrate this machine learning along with all the tools needed to run it (e.g. notifications and integrating this data into other action tools).
They built the systems – still in use and filtered “hundreds of millions of dollars worth of transactions” – but also took an insight in the process: others were surely facing the same issues as them, so why not work on a set of tools to help them all and take some of that work away from them?
Today, Baseten’s main customers – a reference to base ten blocks, often used to help young students learn the basics of math (“It humanizes the number system, and we also wanted to make machine learning less abstract,” the CEO said.) — are developers and data scientists who potentially adopt other machine learning models, or even create their own, but lack the skills to practically integrate them into their own production flow. There, Baseten is part of a larger group of companies that seem to be emerging building “MLops” solutions – comprehensive sets of tools to make machine learning more accessible and usable by those working in devops and developers. products. These include Databricks, Clear, Gathr and more. The idea here is to give tools to technicians to give them more power and more time to work on other tasks.
“Baseten eliminates the process of creating tools so that we can focus on our core competencies: modeling, measurement, and problem solving,” Nikhil Harithras, principal machine learning engineer at Patreon, said in a statement. Patreon uses Baseten to help run an image classification system, used to find content that violates its community guidelines.
Over time, Baseten could take a logical step, continuing its trajectory of democratization: thinking about how to create tools for a non-technical audience as well – an interesting idea in light of the many no-code and low-code products which are deployed to empower them to create their own data science applications as well.
“Non-technical audiences are not something we’re focusing on today, but that’s evolution,” Srivastava said. “The higher goal is to accelerate the impact of machine learning.”