Senior ML Ops Engineer

Be the first dedicated owner of Confido's ML platform. Our AI/ML team already ships document-understanding, forecasting, and agentic systems into production — on infrastructure we've stood up by hand. You'll own that layer: the pipelines, serving, and cloud foundation that turn models and agents into reliable, cost-efficient production systems, at the scale of hundreds of thousands of documents and heavy LLM/VLM workloads.

Own ML pipelines end to end — experimentation to production — and the infrastructure behind training, inference, and agentic workloads

Give the AI/ML team a paved road: reproducible environments and fast paths from prototype to production, so they can try new models and agents without fighting the infra

Stand up the cloud foundation as Infrastructure as Code and the CI/CD that ships ML safely

Serve and optimize inference and forecasting workloads — latency, throughput, and cost — and the data streams feeding them (e.g. turning a heavy synchronous model call into an async, parallelized one)

Own the data interface with data engineering: serve the right data to models and agents, and write their outputs back into the platform's data systems for the rest of Confido to use

Make reliability, observability, security, and privacy the default — and keep model and agent quality measurable in production through online evals and human-in-the-loop review, not just uptime

Required

5+ years in MLOps, ML platform, AI infrastructure, or platform engineering — on production ML systems, not pipelines on paper

You live at the seam of software and infrastructure: equally at home writing production code and standing up cloud infra.

You've driven a real pipeline end to end and can walk through it: the architecture, the security and cost trade-offs, and what you'd change

Deep cloud infrastructure understanding, distributed data systems, and IaC — you can boot an environment from scratch, wire CI/CD, and run containerized workloads in production without hand-holding

Strong Python and comfort in a production app codebase (Ruby, Java) monitoring, security, and cost are instincts, not afterthoughts

High ownership in a fast-moving startup, and experience productionizing what research/AI teams build

LLMOps tooling — tracing, prompt/version management, eval harnesses

Inference optimization (vLLM, ONNX, TensorRT) and GPU / spot-instance economics

ML platform and orchestration tooling (MLflow, BentoML, Ray, Airflow)

Large-scale data systems (Snowflake, Kafka) and vector databases

Managed ML services (Bedrock, SageMaker, Vertex AI)

Multimodal or generative AI in production

Our stack: Python · Ruby/Rails · AWS · Terraform · Kubernetes · GitHub Actions · Snowflake · Aurora/RDS · Redis · Kafka — with more of the above added as we scale.