Baseten

Baseten is a San Francisco-based machine learning infrastructure company that helps developers deploy and serve AI models in production with low latency and reliable scaling.

Baseten was founded in 2019 by Tuhin Srivastava, Amir Haghighat, and Philip Howes, based on the observation that while there had been enormous progress and tooling built around training machine learning models, actually deploying a trained model into a reliable, low-latency, cost-efficient production service that a real application could depend on remained a much harder and more poorly served problem, requiring specialized infrastructure expertise in areas like GPU orchestration, model serving frameworks, and auto-scaling that most application development teams did not have in-house. Baseten's platform provides model inference infrastructure, letting companies deploy open-source or custom AI models, including large language models, image generation models, and speech models, and serve them in production with the reliability, latency, and cost efficiency needed for real customer-facing applications. The company's technology handles the underlying complexity of running AI inference workloads efficiently across GPU hardware, including techniques for reducing latency, batching requests efficiently, and automatically scaling infrastructure up and down based on real-time demand, which is particularly important given how expensive GPU compute capacity remains and how volatile demand for AI-powered features can be. Baseten has increasingly focused on serving generative AI workloads specifically, as demand for reliable, fast open-source model hosting has grown alongside enterprise interest in alternatives to proprietary model APIs from major AI labs, letting companies run open models like those from Meta's Llama family or Mistral with more control over cost, customization, and data privacy than a fully managed proprietary API might offer. The company has raised substantial venture funding from investors including IVP, Spark Capital, and Conviction, reaching a valuation exceeding a billion dollars, competing with other AI inference and model serving platforms including Replicate, Together AI, and the native inference offerings of major cloud providers.

News

No news available for this company.