Python 3.12+ (essential) — Solid production Python, not just scripting level
Snowflake (essential) — comfortable writing and optimizing SQL against Snowflake as a data source and destination for pipeline output
Docker (essential) — building and optimizing container images, including packaging ML model artifacts and their runtime dependencies into deployable images
Kubernetes (essential) — deploying and operating workloads on Kubernetes; understanding pods, deployments, resource limits, and basic cluster networking
Ability to take a research-level prototype and turn it into production-ready, maintainable, scheduled code
Intermediate-advanced or advanced English (indispensable)
Preferred:
Specific experience with Kubernetes CronJob (scheduling, concurrency policy, retry limits, failure history handling)
Experience with KServe model serving platform or equivalent (Seldon, BentoML, or similar), as building a KServe cluster is the next defined phase of this role
Familiarity with clustering/statistics concepts used in the prototype: K-means with auto-K via silhouette scoring, Jenks Natural Breaks for tiering, feature normalization/weighting Authoring Helm charts for packaging Kubernetes workloads
Experience with model artifact formats and packaging conventions (e.g., ONNX, PyTorch/scikit-learn models in pickle, MLflow model registries) even without having trained the models directly
