How Cursor Trained Composer on Fireworks: Distributed Infrastructure for High-Performance RL

How Cursor Trained Composer on Fireworks: Distributed Infrastructure for High-Performance RL

From Training Data by Sequoia Capital

May 26, 2026 · 46 min

About this episode

Federico Cassano and Dmytro Dzhulgakov discuss their collaboration on Composer, a specialized foundation model for software engineering.

Cursor's Federico Cassano and Fireworks' Dmytro Dzhulgakov explain how they collaborated to build Composer as a specialized foundation model. The core insight: models have finite capacity in their weights, and allocating all those bits to the singular task of software engineering in Cursor frees the model to be both better at the task and far more efficient at inference. Rather than start from pre-training and work up, they took an unconventional top-down approach — mid-training and RL on top of an open-source base to get a useful model into users' hands fast, then specializing the model around real Cursor usage. With Fireworks providing distributed infrastructure, Composer delivers frontier-class coding performance with the speed of a much smaller model. Hosted by Sonya Huang, Sequoia Capital

People in this episode

Host: Sonya Huang

Guests: Federico Cassano, Dmytro Dzhulgakov

Topics covered

  • machine learning
  • distributed infrastructure
  • high-performance computing
  • software engineering
  • foundation models

Keywords

  • Composer
  • Cursor
  • Fireworks
  • reinforcement learning
  • distributed infrastructure
  • model training
  • software engineering

Mentioned in this episode

Organizations: Cursor, Fireworks, Sequoia Capital

More episodes of Training Data

Explore listener stats, chart rankings, contacts and more on the Training Data podcast page.