MLOps Engineer

Build the training and deployment infrastructure behind Solace's recommendation models - feature stores, reproducible pipelines, and the on-call rigor a production ML system actually needs.

About Apple

Solace AI builds recommendation and personalization infrastructure for e-commerce platforms. Our models sit directly in the checkout path for several large retail customers, which means model reliability is a revenue problem, not just an accuracy metric. We're a 45-person, Series B company.

About The Role

You'll build and operate the infrastructure connecting our data scientists' training pipelines to production inference - feature store consistency, reproducible retraining, and deployment gates that catch a regressed model before it ever reaches a customer's checkout flow.

Responsibilities

• Own the feature store ensuring training/serving parity across our recommendation models
• Build reproducible, versioned training pipelines (data, code, and model versioning as one unit)
• Implement evaluation gates that block deployment of a model that regresses key metrics
• Operate the online inference service, including latency and availability SLOs
• Partner with data scientists to productionize research code without rewriting it from scratch each time
• Carry on-call for the production inference path (roughly 1 week in 6)

Requirements

• 3+ years building production ML infrastructure, not just training models
• Strong Python and working knowledge of at least one ML framework (PyTorch or TensorFlow)-
• Experience with container orchestration (Kubernetes) for serving ML workloads
• Comfortable owning the reliability of a system data scientists depend on daily

Desirables

• Experience with a feature store platform (Feast, Tecton, or an in-house equivalent)
• Familiarity with data/model versioning tools (DVC, MLflow, or similar)
• Prior experience where a model's failure had direct, measurable business impact

Benefits

• Team currently spans 4 countries - written async communication is a core part of how we work