Baseten
Post-Training Research Engineer
$200K–$275K
San Francisco · Hybrid · Full-time · Posted Mar 23, 2026
About the job
Build in-house tooling that enables efficient, large-scale post-training of diverse model architectures for Baseten customers.
What you'll do
- Develop tooling to support post-training models using varied techniques and architectures.
- Improve the efficiency and quality of training at scale across the software and systems stack.
- Profile and optimize distributed GPU programs and collaborate with researchers to specify and execute solutions.
What they ask for
- Deep understanding of modern ML techniques and tools for training transformers.
- Advanced experience with a tensor or array computation library such as PyTorch, TensorFlow, or Jax.
- Detailed understanding of transformer training parallelism strategies.
- Ability to profile and improve distributed GPU programs and perform roofline analysis.
Nice to have
- Familiarity with HPC and distributed computing platforms such as Slurm, Ray, Kubernetes, and Dask.
- Familiarity with cluster networking technologies such as Infiniband, RoCE, and GPUDirect.
Pay
$200K – $275K USD base salary
Tools
PyTorch, TensorFlow, Jax, Slurm, Ray, Kubernetes, Dask, Infiniband
AI focus
Develop training infrastructure and techniques for post-training models, including reinforcement learning and supervised finetuning.
Model-written summary · the original posting is authoritative