Post-Training Research Engineer

About this role

Baseten seeks a Post-Training Research Engineer to build internal tooling for custom model training using reinforcement learning, supervised finetuning, and novel techniques. You'll work across the full stack—from GPU kernels and distributed systems to Kubernetes infrastructure—enabling customers to deploy highly optimized models in production.

What you'll do

• Develop and maintain tooling for training diverse model architectures with various post-training techniques at scale
• Profile and optimize distributed GPU training programs to improve performance and efficiency
• Implement transformer training parallelism strategies including data, tensor, and pipeline parallelism
• Collaborate with researchers to translate specifications into robust implementations
• Work across systems-level concerns including Kubernetes, storage, and networking topologies
• Contribute to research efforts in model distillation, reinforcement learning, and inference optimization

What they're looking for

• Deep knowledge of transformer training and modern ML techniques
• Advanced PyTorch, TensorFlow, or JAX experience
• Distributed training parallelism (data, tensor, pipeline, context parallelism)
• Performance profiling and roofline analysis for GPU programs
• HPC platforms (Slurm, Ray, Kubernetes, Dask)
• Operating systems fundamentals (processes, kernel, containerization, networking)
• Cluster networking technology (Infiniband, RoCE, GPUDirect)
• Problem-solving with ability to challenge assumptions and approach

Benefits

• Competitive compensation with meaningful equity
• 100% medical, dental, and vision insurance coverage for employee and dependents
• Flexible PTO with company-wide winter break
• Paid parental leave
• Fertility and family-building stipend through Carrot
• Company-facilitated 401(k)