CoreWeave
Software Engineer, Inference
$120K–$160K
Sunnyvale, CA / Bellevue, WA +2 · Full-time · Posted Sep 25, 2026
- Location
- Sunnyvale, CA / Bellevue, WA
Bellevue, WA
Sunnyvale, California, United States - Experience
- 0–3 years of software engineering experience (or equivalent) building cloud/distributed services in production.
- Visa sponsorship
- Position requires access to export-controlled information. Applicants must be a U.S. person, eligible to access the information without export authorization, or eligible and reasonably likely to obtain the required authorization. CoreWeave may decline to pursue export licensing for legitimate business reasons.
About the job
Own and operate a production model-serving service, improving its reliability, observability, latency, and throughput.
What you'll do
- Design, implement, test, deploy, and operate a model-serving service; maintain its dashboards and alarms.
- Improve service performance through micro-batching, concurrency tuning, and caching; automate runbook steps.
- Participate in on-call, lead service-related post-incident actions, and review peers’ code and designs.
What they ask for
- Proficiency in Python or Go; familiarity with gRPC/HTTP APIs and containerized deployments.
- Working knowledge of Kubernetes, Slurm, or Ray and CI/CD pipelines.
- Experience with metrics/alerting tools such as Prometheus, Grafana, or OpenTelemetry, and structured logging.
- Basic understanding of GPU inference concepts, including KV caching, mixed precision, and streaming tokens.
Nice to have
- Experience with inference frameworks such as Triton, vLLM, TensorRT-LLM, Ray Serve, or TorchServe.
- Performance profiling or benchmarking experience, or contributions to SLI/SLO or postmortem practices.
Pay
$120,000.00 - $160,000.00 USD base salary
Tools
Python, Go, gRPC, HTTP, Kubernetes, Slurm, Ray, Prometheus
AI focus
Develop and operate model-serving infrastructure for GPU inference, including performance improvements for generative AI and real-time applications.
Model-written summary · the original posting is authoritative