ML Researcher - Posttraining

Highlights

• Finetune diffusion models at scale to boost image aesthetics and quality.
• Build reward systems and eval suites for reinforcement learning in image space.
• Help advance Krea 2 foundation model built for aesthetic diversity and control.

What they offer

• Generous salary and equity packages
• 100% health and 99% dental/vision premiums covered, FSA and long-term disability
• Flexible PTO policy
• 401k with 4% company match
• Breakfast, lunch and dinner in the office
• Ubers covered to and from the office
• World-class team with significant company-wide impact

Ideal candidate

Researcher passionate about creative AI who thrives in fast, ambiguous environments with strong taste for simple, scalable methods and bias to action.

Requirements

• Proven work posttraining diffusion models for image or video generation
• Experience with large-scale model training, inference, and optimization
• Strong understanding of LLM and diffusion post-training pipelines including PPO, GRPO, DPO, OPD, MOPD
• Strong proficiency in PyTorch and understanding of its inner workings
• Strong background in distributed training (FSDP, CP, SP, USP, TP, EP) and parallelism tradeoffs
• Knowledge of low-precision training/inference in FP8, NVFP4, MXFP8
• Familiarity with fast inference engines (vLLM, sglang) and RL frameworks (slime, verl, etc.)
• Experience training VLM models as reward models
• Comfortable breaking down ambiguous research goals with good research taste