ML Inference Engineer — Scale & Reliability

Clera; San Mateo (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Clera is an early-stage enterprise AI startup hiring an hands-on ML Infrastructure Engineer to own end-to-end inference and model-serving infrastructure. You will build scalable systems enabling AI agents to run reliably under high concurrency, collaborate with ML and infra teams for performance, and optimize latency and throughput in production environments.

You will work on Docker/Kubernetes based deployments with TensorFlow Serving, TorchServe, Triton, and KServe, using observability stacks

Qualifications

• 5+ years of experience building and operating ML inference systems or ML infrastructure in production.
• Hands-on experience with inference-serving frameworks (TensorFlow Serving, TorchServe, Triton, KServe) and custom systems.
• Strong track record optimizing latency, throughput, and reliability at scale.

Responsibilities

• Own inference and model-serving infrastructure end to end, from design through production deployment.
• Build and scale systems enabling AI agents to run reliably under increasing concurrency.
• Collaborate with ML and infra teams to ensure seamless integration and performance optimization.
• Identify infrastructure bottlenecks and drive cross-functional solutions across engineering teams.

Skills

ML inference

Latency optimization

Distributed systems

Observability

Cloud platforms

Backend languages

Tools

Docker

Kubernetes

TensorFlow Serving

TorchServe

Triton

KServe

Prometheus

Grafana

ELK

Neo4j

Job description

Clera is an early-stage enterprise AI startup hiring an hands-on ML Infrastructure Engineer to own end-to-end inference and model-serving infrastructure. You will build scalable systems enabling AI agents to run reliably under high concurrency, collaborate with ML and infra teams for performance, and optimize latency and throughput in production environments.

You will work on Docker/Kubernetes based deployments with TensorFlow Serving, TorchServe, Triton, and KServe, using observability stacks