Research Engineer - RL Infrastructure
Prime Intellect · Remote · Remote ·
COMPENSATION
€150k – €350k
Listed on the employer's job board (Ashby)
- Fully remote or SF
- RL Infrastructure
- PyTorch + DeepSpeed
- Visa sponsorship
- $150k-$350k
Join Prime Intellect to develop the open superintelligence stack, providing infrastructure for post-training at scale. You will work on the Lab platform, unifying compute, environments, and evaluations to support autonomous research and agent workflows.
The team focuses on building systems that allow models to train faster and adapt continuously. You will contribute to the prime-rl framework and collaborate with researchers to translate bottlenecks into concrete system improvements.
What you'll do
- Build and optimize infrastructure for large-scale RL and distributed training
- Improve end-to-end training efficiency across compute and networking layers
- Design low-level performance optimizations for kernels and runtime
- Work on distributed training systems using data, tensor, and pipeline parallelism
- Shape the architecture of RL training stacks and async rollout systems
- Contribute to open-source libraries for frontier-scale model training
- Collaborate with researchers to identify and resolve system bottlenecks
What you'll need
- Strong systems engineering experience in AI/ML infrastructure
- Deep familiarity with PyTorch and distributed training frameworks
- Experience optimizing training performance across kernels and memory
- Hands-on experience with large-scale training parallelization techniques
- Strong understanding of GPU architecture and performance debugging
- Ability to drive improvements from first principles in ambiguous contexts
Nice to have
- Experience writing or optimizing CUDA or Triton kernels
- Background in compiler or runtime optimization for ML systems
- Experience with RL training infrastructure or async pipelines
- Work with multi-node GPU clusters and high-performance networking
- Contributions to open-source ML systems or infrastructure projects
Stack
- PyTorch
- DeepSpeed
- FSDP
- Megatron
- vLLM
- Ray
- CUDA
- Triton
How you'll work
- Flexible work arrangements available
- Option to work remotely or in San Francisco
- Visa sponsorship and relocation assistance provided
What you get
- Cash compensation range of $150,000 to $350,000
- Equity incentives aligned with company growth
- Quarterly team off-sites and hackathons
- Conference attendance and learning opportunities