Machine Learning Performance Engineer

**About Jane Street:** Jane Street is a quantitative trading firm that leverages cutting-edge technology to develop elegant solutions that scale well. Our core belief is that building robust and reliable applications should not require developers to manage complex state or worry about underlying infrastructure. We are committed to a user-obsessed culture, emphasizing employee well-being with competitive salaries, equity options, and comprehensive benefits. Our remote-first culture with flexible schedules allows us to collaborate effectively across time zones and locations.

**About The Role:** We are looking for an engineer with experience in low-level systems programming and optimization to join our growing ML team.

**Key Responsibilities:**
- Optimize the performance of machine learning models for both training and inference
- Improve CUDA and other low-level systems for ML operations
- Debug performance issues in ML training and inference pipelines
- Enhance storage systems, networking, and host-level considerations for ML
- Ensure efficient use of resources and high-throughput processing

**Required Skills and Ideal Candidate:**
- Deep understanding of modern ML techniques and toolsets
- Experience debugging end-to-end performance in training runs
- Knowledge of low-level GPU programming (PTX, SASS, Tensor Cores)
- Proficiency with optimization tools like CUDA GDB, NSight Systems
- Familiarity with libraries such as Triton, CUTLASS, CUB
- Intuition about CUDA performance characteristics
- Background in networking technologies (Infiniband, RoCE)
- Understanding of distributed training algorithms (NCCL, MPI)
- In inventive approach to solving complex performance problems