Jordan Blake

Jordan Blake did not set out to work on AI. They set out to build things that worked. After several years at Stripe building payment infrastructure, they moved to Uber ATG where they encountered machine learning at scale for the first time — and found it fascinating, frustrating, and unlike anything they had built before.

At Cohere, they became one of the earliest engineers working on practical LLM deployments outside of research settings. They built internal tooling for evaluation, fine-tuning pipelines, and the first production retrieval-augmented systems the company shipped. Their GitHub contributions to open-source AI tooling have tens of thousands of stars.

At LangChain, they work on making it easier for developers to build reliable AI applications. They are known in the community for being unusually honest about what does and does not work — and for writing blog posts with titles like 'When Not to Use an LLM' and 'The 10 Ways Your Agent Will Fail in Production.'

Jordan does not have a PhD and considers this mostly irrelevant. 'The field moves too fast for credentials to matter much,' they say. 'What matters is whether you can ship something that works tomorrow.' They mentor dozens of engineers transitioning into AI through open office hours and a popular YouTube channel.