Company
Decart
Annual salary
$350k – $450k/yr
Location
San Francisco, CA
Listed
29d ago
- Experience:
- 5+ years
- Workplace:
- 5 days in-office in San Francisco, CA
- Equity:
- Competitive equity
Open to people who live in the US, Canada or the UK
We can only put candidates forward for this role from those three countries for now. See the contract AI training jobs open in more countries.
On-site in San Francisco, CA.
Send us your LinkedIn and CV. If your experience fits, we'll introduce you to the recruiter filling this role.
Apply for a referral →The recruiter emails you before anything happens and may suggest other jobs that suit you better.
About this role
You'll diagnose and resolve performance problems across Decart's ML systems, spanning research, training, and production inference. The largest share of the work is writing and optimizing kernels for TPU and Trainium. You'll also advise researchers on the performance cost of proposed model changes. We're looking for engineers with a demonstrated record in large-scale systems engineering and low-level optimization.
What you need
- Bachelor's degree in Electrical/Computer Engineering, Computer Science, or a related field, plus 5+ years of relevant experience (or equivalent practical experience)
- Solid grounding in operating systems fundamentals (process/thread scheduling, synchronization, virtual memory), CPU/GPU architecture, and hardware/software co-design
- Working knowledge of PyTorch and machine learning algorithms, with a focus on engineering application
- Demonstrated ability to profile compute and memory behavior, diagnose bottlenecks, and validate improvements with rigorous measurement
- Production experience squeezing performance out of ML workloads on TPU, Trainium, GPU, or other accelerators
- You've authored kernels for an ML accelerator, not just consumed them
- Depth in computer architecture: you can reason about systolic arrays, memory hierarchies, and interconnect topologies from first principles
- Familiarity with compiler and toolchain internals (e.g., XLA, MLIR, Triton, or vendor stacks)
- Experience scaling training or inference workloads across multi-accelerator clusters
- You've read, or patched, the internals of an ML framework
Projects you might work on
Cut milliseconds off end-to-end token latency in DOS by restructuring attention and sampling paths for a new accelerator generation
Design communication schedules that overlap compute with network transfer across multi-chip topologies
Build analytical performance models to predict where the next 2x is hiding before writing a line of kernel code
Trace a throughput regression from a framework-level symptom down to instruction scheduling in generated assembly, and fix it
Port DOS's kernel suite to new hardware and close the gap to theoretical peak
Common questions
How do I apply for the Kernel Engineer (ML Accelerators) (US) role at Decart? +
Use the Apply for a referral button on this page to send us your LinkedIn and CV. If your experience fits, we'll introduce you to the recruiter filling this role. They'll email you to check you're interested, then put you forward for this job or others that suit you better. It's free.
What does this Decart role pay? +
The listing gives $350k – $450k/yr.
Is this role remote? +
The listing gives the location as San Francisco, CA. 5 days in-office in San Francisco, CA.
Related roles
Browse all referral roles
Salaried roles at startups, AI companies and established businesses, all filled by referral. One application covers every role.
See all referral roles →