Baseten Labs Inc., a company dedicated to simplifying developers’ interaction with artificial intelligence concepts in production, has recently concluded a \(40 million funding round. The funding was spearheaded by IVP and Spark Capital, with participation from previous sponsors. Forbes now values Baseten at over \)200 million following the Series B raise.
The process of running AI models in production demands significant time and effort to guarantee optimal performance. Developers emphasize the necessity of preparing the model’s operating infrastructure for unexpected surges in traffic, ensuring timely responses to customer requests, averting excessive cost escalations, and managing various technical tasks.
Headquartered in San Francisco, Baseten aims to streamline this process. From the initial model implementation phase onwards, the company provides a platform designed to handle many of the tasks associated with operating production Artificial Intelligence environments.
Developers are tasked with converting a newly trained neural network into a format compatible with their team’s cloud infrastructure. Baseten’s open-source tool, Truss, purportedly expedites this process, enabling the deployment of AI models on its platform with minimal code input.
To expedite consumer request processing when a neural network is active, Baseten employs a policies engine. This engine generates replicas of the AI model and redistributes increased traffic among them upon detecting a surge in usage, ensuring prompt responses.
Furthermore, Baseten automates the removal of replicas once user activity normalizes. Developers can leverage the platform to implement a “scale to zero” approach, wherein an AI workload halts completely during periods of inactivity, thus mitigating cost escalations.
By utilizing Baseten’s dashboard, developers can monitor resource consumption of AI models and key metrics such as request processing times. The company also offers an observability tool to facilitate the troubleshooting of complex issues.
In a recent blog post, CEO Tuhin Srivastava highlighted that Baseten’s native workflows facilitate the seamless deployment of large models in production, eliminating the need for users to manage version control, deployment, and observability. The platform has reportedly scaled production loads significantly without downtime.
Baseten’s platform is adaptable for businesses as a managed cloud service or as a self-hosted application within Google Cloud and Amazon Web Services environments. With approximately 20 large enterprises and tens of thousands of developers utilizing its platform, Baseten’s quarterly revenue is estimated to be in the “mid single digit millions.”
Looking ahead, the company plans to expand its sales and marketing team nearly threefold by year-end to bolster its client base. The recently secured funding will fuel product development initiatives, focusing on enhancing support for cloud platforms and tools that optimize customers’ AI models’ performance and streamline training processes.
