Staff ML Engineer - Post-Training & RL Pipelines

SF Tensor; San Francisco (CA); On-site; USD 275,000 - 315,000; Full time

Job summary

SF Tensor is building the fastest GPU compiler and a high-performance enterprise modeling platform in San Francisco. This role centers on owning the post-training pipeline end-to-end—from data curation through SFT, RL, distillation and deployment.

You'll work with customers’ domain experts to craft reward functions, design evaluation frameworks and build an eval harness that supports ship/no-ship decisions in a tight window.

Qualifications

• Experience shipping post-trained models into production.
• Hands-on depth with SFT and RL (DPO, GRPO, PPO or similar).
• Ability to design evaluation metrics and interpret results honestly.
• Experience managing data curation, labeling workflows and synthetic data generation.

Responsibilities

• Own the post-training pipeline end-to-end including data curation, SFT, RL, evals and distillation.
• Design reward functions and RL objectives with customer domain experts.
• Build an eval harness capable of ship/no-ship decisions within tight windows.
• Structure datasets including synthetic data pipelines from customer data.
• Distill specialist models into smaller, deployable versions.
• Identify critical path bottlenecks and optimize iteratively.
• Collaborate with customers as a forward-deployed researcher and transfer ownership.

Skills

Post-training ML

Reinforcement learning

Evaluation design

Data curation

Tools

PyTorch

JAX

Job description

SF Tensor is building the fastest GPU compiler and a high-performance enterprise modeling platform in San Francisco. This role centers on owning the post-training pipeline end-to-end—from data curation through SFT, RL, distillation and deployment.

You'll work with customers’ domain experts to craft reward functions, design evaluation frameworks and build an eval harness that supports ship/no-ship decisions in a tight window.