OpenPipe
Help developers and enterprises build reliable, high-performance AI agents through specialized fine-tuning and reinforcement learning.
At a Glance
- AI Software Developers
- Enterprise Data Science Teams
- Cloud-Native AI Startups
AI Tools by OpenPipe
(1)
ART (Agent Reinforcement Trainer)
RL Training Framework for LLM Agents
Discussions
No discussions yet
Be the first to start a discussion about OpenPipe
Latest News
OpenPipe training and inference migrating to Weights & Biases (W&B)
Launch of Serverless RL for faster and cheaper model training
CoreWeave enters definitive agreement to acquire OpenPipe
OpenPipe raises $6.7M Seed round led by Costanoa Ventures
Products & Services
A managed service for supervised fine-tuning (SFT) to migrate generic prompt chains to specialized models.
Tools to align AI agents with specific business objectives through iterative reinforcement learning.
An open-source library and framework for training autonomous AI agents.
Pay-as-you-go infrastructure for training LLMs with reinforcement learning.
Market Position
Specializes in moving generic prompt chains to specialized, smaller models using RL, offering better performance at lower cost compared to Together AI or Anyscale.
Leadership
Founders
Kyle Corbitt
Co-founder & CEO. Previously Director of Startup School at Y Combinator. Software engineer at Google and Meta. Founder of Emberall.
David Corbitt
Co-founder & CPO. Previously Software Engineer at Qualtrics and Palantir. Co-founder of GenerationalStory.
Executive Team
Kyle Corbitt
Co-founder & CEO
Former Director of YC Startup School and engineer at Google and Meta.
David Corbitt
Co-founder & CPO
Former engineer at Palantir and Qualtrics; leads product strategy.
Board of Directors
Founding Story
Founded by brothers Kyle and David Corbitt in 2023 to address the high costs and latencies of using large LLMs by enabling developers to train smaller, specialized models that 'learn from experience'.
Business Model
Revenue Model
Usage-based API fees (per token for training and inference) and hourly compute unit billing.
Pricing Tiers
Supervised fine-tuning for models with ~8B parameters.
Supervised fine-tuning for mid-sized models.
Supervised fine-tuning for large scale models like Llama 3 70B.
Input/Output pricing for specialized hosted inference.
Target Markets
- AI Software Developers
- Enterprise Data Science Teams
- Cloud-Native AI Startups
- Autonomous Production Agents
- Deep Research Bots
- Enterprise KPI Optimization
- ART.E
- Autonomous Agent developers