Lead design and delivery of high-scale, resilient distributed inference data plane services. Architect and optimize inference hosting (tensor/data parallelism, KV-cache routing, prefill/decode disaggregation), contribute upstream open-source projects, mentor engineers, and maintain operational excellence and SLOs for a multi-tenant AI inference cloud.

Lead architecture and low-level optimizations for high-performance AI inference. Benchmark and tune GPU kernels, memory and precision management, parallelize across multi-node GPU clusters, implement quantization (FP8/INT8/FP4), advise on hardware/software stacks (CUDA/ROCm/Triton/TensorRT), mentor engineers, and collaborate to turn hardware limits into shippable inference products.