Case Study: Replit achieves faster AI model development and time to market with Databricks

Empowering the next billion software creators

Replit, a software startup with a platform that helps developers complete, explain, and transform code, needed a way to train a specialized code-completion model quickly so it could scale GenAI-powered software development and meet an accelerated product launch timeline. To avoid the complexity of building end-to-end training infrastructure itself, Replit turned to Databricks and its Mosaic AI Training product.

Using Databricks Mosaic AI Training, Replit built its own code completion model from scratch in less than a week, helping it launch on time with fewer engineering headaches. Databricks also helped boost the productivity of Replit’s AI engineers, speeding time to market for new models and delivering significant business impact.

Replit

Michele Catasta

VP of AI