Announcing Vellum

Hi everyone 👋 ‍ We’re excited to publicly announce the start of our new adventure: Vellum. We’re in Y Combinator’s current batch (W23) and our mission is to help companies get the best results from Large Language Models like GPT-3. Our product helps developers evaluate, manage and A/B test AI models/prompts to increase quality and reduce cost. Since GPT-3 launched in 2020 we saw companies like Jasper and find compelling sales & marketing use cases. In the last 2 years the rate of improvement of these foundation models has been staggering, as clearly evidenced by OpenAI’s ChatGPT and models from Cohere and AI21. With all these advances, companies around the world are looking to incorporate Large Language Models (LLMs) for generation and classification use cases both for internal applications and in their core product. However, we’ve seen 3 challenges when companies try to bring these models into production. These obstacles result in slower iteration cycles and suboptimal configurations

Hi everyone 👋 ‍ We’re excited to publicly announce the start of our new adventure: Vellum. We’re in Y Combinator’s current batch (W23) and our mission is to help companies get the best results from Large Language Models like GPT-3. Our product helps developers evaluate, manage and A/B test AI models/prompts to increase quality and reduce cost. Since GPT-3 launched in 2020 we saw companies like Jasper and find compelling sales & marketing use cases. In the last 2 years the rate of improvement of these foundation models has been staggering, as clearly evidenced by OpenAI’s ChatGPT and models fr