Baseten
Production AI inference platform for custom and open-source models.
Description
ššš¬šššš§ is an AI inference platform founded in 2019 by Tuhin Srivastava, Amir Haghighat, Philip Howes, and Pankaj Gupta after the team experienced model deployment problems firsthand at Gumroad. Headquartered in San Francisco, the platform handles dedicated GPU serving, multi-cloud routing, and autoscaling for AI products in production. Customers include Abridge, Writer, Clay, OpenEvidence, and Descript, with healthcare and voice AI representing two of the strongest verticals.
Key Capabilities:
ā ššš¬šššš§ Inference Stack with custom kernels and advanced caching
ā Cross-cloud high availability across multiple regions and providers
ā Self-hosted and VPC deployment options for single-tenant workloads
ā Truss open-source framework for ML model packaging
ā Pre-optimized API access for Llama, Mistral, DeepSeek, and Whisper
ā ššš¬šššš§ Chains for compound AI with granular hardware control
ā ššš¬šššš§ Embeddings Inference for vector workloads
ā Training pipeline with one-click deployment to inference infrastructure
ā Auto-scaling with scale-to-zero for idle workloads
ā Cold start optimization with 99.99% uptime target
ā Forward deployed engineering support for production rollouts
ā Python SDK and CLI for programmatic deployment control
Alternative tools
- Modal
Serverless GPU platform for AI inference, training, and batch jobs.
- Fly.io
Run full-stack apps and databases on global microVMs.
- DeployLens
Real-time visibility into your GitHub deployment pipeline.
