Senior Applied ML Engineer | Kimchi (Model Router)

Kimchi is hiring a Senior Applied ML Engineer to build the model that chooses which LLM serves each request. The role focuses on predicting quality, cost, and decoding speed from production and benchmark data, then improving routing decisions while maintaining task quality. You would own the model’s features, objective, training, calibration, evaluation, and releases.

The work spans training signal from fleet records and benchmark rollouts, production-parity evaluations with error bars, and online experiments from shadow runs through A/B tests and full rollout. The role also calls for monitoring drift, investigating differences between offline evaluations and production, and partnering with platform engineers on instrumentation. Candidates need 5+ years in applied ML or ML research and experience shipping real-time decision models. The listing describes a remote-first global environment and names Italy among eligible locations.

Highlights

• Own the routing model end to end, from training to release.
• Build rigorous evaluations and turn fleet outcomes into training data.
• Run online experiments, monitor drift, and retrain.
• Work with Python, PyTorch, SQL at scale, and LLM APIs.