Model Serving Engineer

Requirements

• Design and operate model serving platforms supporting diverse workloads including LLMs, vision models, and recommendation systems.
• Optimize inference performance using continuous batching, paged attention, speculative decoding, and request multiplexing.
• Implement multi-tenant routing, rate limiting, and quality-of-service policies across model endpoints.
• Build autoscaling and capacity management systems that balance latency, throughput, and cost.
• Tune GPU utilization, memory management, and KV cache strategies for LLM serving workloads.
• Integrate model serving with API gateways, identity systems, and observability platforms.
• Implement caching, prompt deduplication, and response reuse strategies where appropriate.
• Drive end-to-end observability including latency histograms, queue dynamics, GPU utilization, and error tracking.
• Develop deployment workflows including canary releases, shadow testing, and automated rollback.
• Operate incident response for high-availability AI services and drive durable reliability improvements.
• Collaborate with ML and product teams to support new model releases and capability rollouts.
• Implement security controls including request signing, content filtering, and abuse detection at the serving layer.
• Document operational procedures, performance characteristics, and tuning guidance for internal teams.
• Stay current with AI serving research and translate advances into production capabilities.

Benefits

• Competitive base salary commensurate with experience, plus benefits.