Build the foundational inference cloud platform, including its control plane, API, serving layer, routing, model placement, scheduling, reliability, and observability. The role involves scaling a heterogeneous GPU and ASIC fleet, partnering with data center and model deployment teams, and shaping platform architecture as an early individual contributor at a small startup.

Own the end-to-end LLM inference serving stack for specialized ASIC hardware. Design request routing, batching, scheduling, KV-cache management, autoscaling, monitoring, alerting, and failover. Optimize throughput and cost per token, support production reliability and on-call operations, collaborate with compiler and hardware bring-up teams, shape the serving roadmap, and establish technical standards for a growing engineering team.