Engineering Manager - Inference Performance

We're looking for an Engineering Manager to lead part of our Inference Performance team. This team makes the world's most demanding AI workloads run faster and more efficiently on GPUs. You'll manage and grow a team of inference performance engineers working across the inference engine and runtime: kernels, scheduling, batching, KV-cache management, speculative decoding and prefill/decode disaggregation. This is a hands-on technical leadership role. You'll set direction, unblock hard problems and earn the team's trust by going deep on GPU performance, while also hiring, developing and supporting the people doing the work. Your team's output directly affects how fast our customers' models run and how efficiently we serve them. The team is scaling quickly, so you'll help shape how it is structured as it grows.

Your team will work on these types of projects as part of our Inference Runtime team:

• Agentic inference optimization: 50-90% faster engines https://www.baseten.co/blog/agentic-inference-optimization-faster-than-sota/
• Agentic Kernels in Production https://www.baseten.co/blog/agentic-kernels-in-production/
• Live draft model training for speculative decoding https://www.baseten.co/blog/live-draft-model-training-for-speculative-decoding/
• The Baseten Inference Stack https://www.baseten.co/resources/guide/the-baseten-inference-stack/
• Lead, mentor and grow a team of inference performance engineers through regular 1:1s, clear feedback, career development and performance reviews.
• Hire top GPU and inference engineering talent, and build a strong, collaborative team culture as the runtime team scales.
• Own the technical roadmap and execution for runtime performance work, balancing customer needs, new model launches and long-term platform investments.
• Stay close to the technical work. Review designs, guide profiling and optimization efforts, and help the team reason from first principles about where time and memory go.
• Drive the productionization of inference techniques such as quantization, speculative decoding, KV-cache reuse, chunked prefill and custom scheduling.
• Turn performance wins into measurable outcomes: tokens per GPU-hour, utilization, latency and cost.
• Help the team bring up and tune new model architectures on new hardware quickly, often in the same week they're released.
• Partner with Infrastructure, Inference Platform, Kernels, Model APIs and customer-facing teams to set priorities, coordinate launches and ship wins.
• Set high standards for engineering quality, benchmarking, operational excellence and incident response.
• Bachelor's, Master's, or Ph.D. degree in Computer Science, Engineering, Mathematics, or a related field.
• Experience managing engineers, including hiring, mentoring, giving feedback and running performance reviews.
• Experience leading or closely supporting GPU optimization teams in training, inference or recommendation systems.
• Strong technical depth in GPU workloads, with a solid understanding of GPU architecture and performance tradeoffs.
• Familiarity with ML libraries such as PyTorch, TensorRT or TensorRT-LLM.
• A track record of driving roadmaps and shipping complex technical projects with a team.
• Clear written and verbal communication, including the ability to align stakeholders across teams.
• Familiarity with inference engines such as vLLM, SGLang or TensorRT-LLM.
• Experience with LLM optimization techniques (e.g., quantization, speculative decoding, continuous batching) in production.
• Experience with GPU kernels (CUDA, Triton, CUTLASS, or similar).
• Experience scaling a team through rapid growth at a startup.
• A background as a hands-on performance or systems engineer before moving into management.
• Paid parental leave