Machine Learning Engineer, Electra

• Strong applied-ML / modelling background: training, evaluating, and optimizing production ML models, with solid ML fundamentals
• Strong software engineering in Python, with production MLOps experience: taking models to production, deployment, monitoring, retraining, model registries
• Comfortable across the data stack: SQL, dbt, Snowflake (or equivalent)
• You operate production systems and care about reliability, reproducibility, testing, CI/CD, and observability
• Bias for shipping. You'd rather have a simple model live and delivering value than a perfect one stuck in a notebook. You ship, measure, and iterate
• Versatile and autonomous. Comfortable taking a model from an early prototype through to production on your own, then moving on to the next use case
• Breadth over depth, by choice. Energised by covering many use cases rather than perfecting a single one. You context-switch well across problems and domains
• Pragmatic about tooling. You build just enough infrastructure to make the next model easier to ship, never a platform for its own sake
• Ownership beyond the model. You care what happens after deployment: reliability, monitoring, and the actual business outcome, not just offline metrics
• Collaborative. You partner closely with the Data Scientist, Analytics Engineers, SRE, and backend teams, and pick things up wherever they need to move
• Even if your profile doesn't match 100% of every criterion, we encourage you to apply. At Electra, we strive to create an environment where everyone feels comfortable, safe, and free to be themselves

Desirable

• Take models from prototype to production. Collaborate with the Data Scientist early on, then own the modelling that makes it production-ready: feature engineering, tuning, training pipelines, and performance optimization
• Own model performance in production: monitor drift, quality, and prediction-vs-actuals (Evidently or similar), set up automated retraining, and iterate on the models themselves when they degrade
• Build the evaluation framework that gates every model rollout, especially for our GenAI use cases (example: ElectraDrive)
• Define a standard, repeatable path from prototype to production for batch and event-driven inference: the deployment contract (standard packaging + model registry) integrating with Dagster (batch) and Temporal (event-driven, backend-owned)
• Set up just enough shared tooling to keep modelling consistent and repeatable: a feature store (dbt/Snowflake) for training/serving consistency and a model registry with versioning and lineage (MLFlow or similar)
• Contribute to the parts of our data-platform infrastructure directly tied to ML workloads: primarily orchestration (Dagster) and, where it supports ML data needs, the lakehouse, alongside the Analytics Engineers and SRE team