Staff MLOps Engineer

What you will do

• Define the target-state MLOps architecture for Elliptic, covering model training pipelines, serving infrastructure, monitoring, feature management, and governance, and produce the architecture decision records that inform investment decisions
• Make and document build-vs-buy-vs-stop recommendations with clear cost modelling and trade-off analysis, evaluating vendors, open-source tools, and managed services against Elliptic's constraints (AWS-primary, Databricks ecosystem)
• Work with InfoSec to improve the existing model registry and model risk management framework, closing identified gaps in metadata, lineage, approval workflows, and drift/bias detection
• Build model training pipelines, CI/CD for ML, and serving infrastructure, working directly with a small group of infrastructure engineers to ship production-grade platform capabilities
• Instrument observability across the ML lifecycle: training metrics, serving latency and throughput, data quality, and prediction drift, integrating with Elliptic's existing observability stack
• Work directly with data scientists and ML engineers across all four consumer groups to onboard them onto the platform, writing documentation, runbooks, and reference architectures that lower the barrier to self-service

You will be a great fit here if you

• Have built MLOps platforms or ML infrastructure from the ground up, and can speak to what worked, what didn't, and why
• Have operated in a regulated industry (e.g. compliance, financial) and have hands on experience building ML infrastructure to meet those regulatory demands
• Think about ML infrastructure the way the best platform engineers think about data infrastructure: as a set of foundations with internal customers whose needs must be understood and balanced
• Are comfortable operating in ambiguity, making decisions with incomplete information, and creating structure where none exists, while remaining open to changing course when better information arrives
• Influence through clarity, evidence, and the quality of your work rather than positional authority. You earn adoption by making the platform genuinely better than the alternative
• Care about production engineering quality: you write production-grade code, your systems are tested, observable, documented, and designed for others to operate

Our ideal candidate has

• Deep hands-on experience building MLOps platforms, including model registries, feature stores, and ML pipeline orchestration
• Working knowledge of model serving patterns: real-time inference, batch prediction, A/B deployment, and deployment strategies
• AWS infrastructure experience (ECS/EKS, S3, IAM, networking) and comfort operating in a Databricks ecosystem or equivalent lakehouse architecture
• Experience with model monitoring: model evaluation, data drift detection, prediction drift, and performance degradation alerting
• A track record of building something from zero and bringing it to a state where others could operate and extend it
• Experience in a regulated industry (fintech, financial services, healthcare) where model governance is a compliance requirement
• See AI as a core part of how modern engineering gets done, not a passing trend. You actively use it to think faster, prototype faster, and pressure-test your own designs, and you're excited that the bar keeps rising.
• Prior experience running formal build-vs-buy evaluations with written decision records

Bonus Points for

• Familiarity with model risk management frameworks and the ability to connect governance practices to regulatory expectations
• Experience working simultaneously with research-oriented ML teams and production-oriented engineering teams, and understanding how their needs diverge
• Infrastructure-as-code fluency (Terraform)
• Experience with ClickHouse or similar OLAP engines for low-latency ML feature serving
• Blockchain or crypto domain knowledge
• Experience working in fraud detection and modelling
• Contributions to open-source MLOps tooling