Member of Technical Staff Kernels
Magic is an AI research and engineering company building frontier code models to automate software engineering and research.
Funding history
About Magic
Magic AI, Inc. develops long-context foundation models and agentic systems for software engineering, combining pre-training, reinforcement learning, ultra-long context, and inference-time compute in pursuit of safe AGI.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will design, implement, deploy, and maintain high-performance kernels that improve throughput and latency for training and inference. You will optimize long-context workloads, test for correctness and robustness, evaluate alternative hardware, and collaborate with training, inference, and reinforcement learning functions.
Requirements
- Low-level programming experience targeting AI accelerators such as NVIDIA Blackwell or Google TPUs
- Experience developing and optimizing GPU kernels
- Deep expertise in computer architecture, low-level machine optimization, and code generation
- Experience with kernel authoring frameworks such as Pallas, Mosaic, or Mojo
Responsibilities
- Design and implement kernels for high-performance long-context behavior
- Own kernel design, implementation, deployment, and production reliability
- Test kernels for robustness and functional correctness while improving performance
- Evaluate porting compute kernels to alternative hardware
- Co-design kernels with training, inference, and reinforcement learning functions
Benefits
- Equity
- 401(k) plan with 6% salary matching
- Health insurance
- Dental insurance
- Vision insurance
- Unlimited paid time off
- Visa sponsorship
- Relocation stipend