OpenPipe, a Seattle startup that wants to make it easier and cheaper for companies to train and deploy large language models, announced a $6.7 million seed round.
The company, part of Y Combinator’s summer cohort last year, lets developers build LLM models tailored to their specific use case. The idea is that smaller models can result in better performance and lower cost compared to using larger general purpose models such as OpenAI’s ChatGPT.
OpenPipe is betting that companies interested in using generative AI might not need a chatbot that can answer any question on any topic — but rather something that understands their product line and company policies deeply to provide customer service or other capabilities.
For example, OpenPipe cited a financial services company that was using OpenAi to process call transcripts and extract information like credit card balances. It was able to significantly reduce cost and the number of errors after switching to OpenPipe.
“Customers end up paying much less than their previous OpenAI bill while benefiting from the higher-quality responses a model tailored to their specific use case produces,” said OpenPipe CEO Kyle Corbitt.
OpenPipe launched last year and said it has “significant growth” but declined to share revenue metrics. It makes money by charging customers for both fine-tuning their models and using the models in production.
Companies across various industries are ramping up their use of generative AI, but the high cost of training, deploying, and updating models has been cited as a potential challenge.
“We’ve found that in practice most businesses are using generative AI for specific narrow use cases within their product or service, which is a great fit for these smaller specialized models,” Corbitt said.
Part of OpenPipe’s pitch is that its software doesn’t require advanced machine learning and data science skills.
“Our users are fullstack app engineers, and they’re able to successfully train really strong models for their specific use cases without involving an outside expert,” Corbitt said.
Corbitt previously founded a family history startup called Emberall before joining Y Combinator, where he was an engineer for the startup accelerator and led Startup School, its founder community.
He co-founded OpenPipe with his brother, David Corbitt, a former engineer at Qualtrics and Palantir who co-founded a video legacy startup called GenerationalStory.
Costanoa Ventures, an early stage firm based in Silicon Valley, led the seed round, which included participation from Y Combinator and individual backers such as Logan Kilpatrick, former head of developer relations at OpenAI, Alex Graveley, creator of GitHub Copilot, and Tom Preston-Werner, co-founder of GitHub.
