Systems Engineer (Inference)

Sapience AI Corporation•Seattle, WA

3d•$204,000 - $216,000

About The Position

The AI/ML Infrastructure and Systems Engineer owns how Sapience AI serves the intelligence behind MINERVA and COGENT in production. The role builds and operates inference infrastructure that delivers fast, reliable, and affordable answers to members, including the end-to-end serving path from request to response. The engineer is responsible for inference performance, latency, scaling, reliability, GPU and hardware efficiency, model deployment and lifecycle, observability, and operational tooling. They make platform choices about accelerators, runtimes, and serving frameworks, balancing performance, cost, and maintainability. The role partners closely with Applied AI, Neuro-Symbolic AI, Platform Engineering, and Research to move models and reasoning components safely from research into production. This is a production infrastructure role focused on inference systems, performance, and hardware efficiency—not model architecture research, general backend engineering, or data engineering. The engineer is expected to use AI to accelerate infrastructure work while retaining ownership of correctness, reliability, cost, and technical judgment.

Requirements

Five or more years of experience in infrastructure, systems, or ML infrastructure engineering.
Hands-on experience serving ML or LLM models in production at scale.
Deep understanding of latency, throughput, and performance optimization.
Experience with GPUs or accelerators and using them efficiently.
Strong systems programming and distributed systems fundamentals.
A track record of building reliable, cost-aware production systems.
Fluency with observability and operational excellence.
Experience with inference-serving frameworks and model runtimes, such as vLLM, TensorRT, or Triton-class systems.
Experience with GPU tooling, CUDA-class ecosystems, and accelerator runtimes.
Experience with Kubernetes, containers, and cloud platforms such as AWS, GCP, or Azure.
Experience with observability stacks for metrics, tracing, and logging.
Python proficiency and experience with a systems language such as Go, Rust, or C++.
Experience with caching, queuing, and load-management systems.
Prior ML infrastructure, platform, or systems engineering experience at a software or AI company.
Experience serving models in production under real load.
A track record of performance and reliability improvements at scale.

Nice To Haves

Experience with inference-serving frameworks and model runtimes.
Experience optimizing LLM inference, including batching, quantization, and caching.
Familiarity with retrieval systems and their performance characteristics.
Experience owning cost and capacity for AI workloads.
Exposure to neuro-symbolic or agentic systems in production.
Experience owning inference cost.

Responsibilities

Build and operate inference systems serving the models and reasoning behind MINERVA and COGENT.
Own the serving path end to end, from request to response, under real production load.
Make serving robust to failures, traffic spikes, and change.
Reduce latency and maintain low response times as the platform grows.
Optimize model execution, batching, caching, and retrieval across the full serving path.
Profile systems and remove significant performance bottlenecks.
Scale inference across more members, communities, and heavier reasoning while maintaining reliability.
Build autoscaling, load management, and graceful degradation capabilities.
Own serving-layer reliability.
Improve GPU and hardware utilization and reduce inference waste without compromising quality or speed.
Manage inference economics and make cost a deliberate property of the system.
Build safe paths for taking models and reasoning components from research into production.
Support confident model rollout, versioning, and rollback.
Instrument the serving layer and build alerts, dashboards, and operational tooling.
Reduce the operational burden of running AI inference at scale.
Make platform decisions about accelerators, runtimes, and serving frameworks, balancing performance, cost, and maintainability.
Keep the inference technology stack current.
Use AI to generate tooling, reason about performance, and accelerate infrastructure work while owning correctness, reliability, and cost.
Collaborate with Applied AI, Neuro-Symbolic AI, Platform Engineering, and Research to serve MINERVA and COGENT models and reasoning in production.