Espresso AI

Espresso AI builds a neural compute optimizer that applies machine learning to performance engineering, running customers' data warehouses in the manner of a team of data engineers working continuously. The company's platform uses models that understand code to optimise individual SQL queries on the frontend while allocating physical compute on the backend. The stated aim is to reduce spending on Snowflake and Databricks without requiring customers to change their workflows.

Deployment is deliberately simple: setup takes about ten minutes, consisting of running a SQL command and adjusting some configurations. The company reports savings of up to 70% on Snowflake and Databricks bills. Technical work spans machine learning, large language models, code intelligence, SQL query optimisation, neural scheduling and compute resource allocation.

Espresso AI is a small team with backgrounds from Google, Apple and MIT. The company maintains a research programme aimed at building a model capable of superhuman code optimisation, and frames its mission as automating performance engineering to deliver improvements not seen since the end of Moore's law. Its customers are organisations running data warehouses on cloud data platforms.