MLOps Engineer

Build infrastructure for training, deploying, and monitoring ML models in production

Manage model versioning, experiment tracking (MLflow, Weights & Biases)

Build feature stores and data pipelines for ML training workloads

Set up model serving infrastructure (Triton, TorchServe, BentoML, FastAPI)

Monitor model drift, data drift, and performance degradation in production

Manage GPU clusters for training jobs (SLURM, Ray, Kubernetes + GPU operators)

Bridge between data scientists and platform/infrastructure teams

Python: expert-level (production-grade OOP, async, packaging, testing)

DevOps/Cloud baseline: Docker, Kubernetes, CI/CD — non-negotiable foundation

ML fundamentals: understand model training, evaluation, overfitting

MLflow / Weights & Biases / DVC for experiment tracking and model versioning

Feature stores: Feast, Tecton, or Hopsworks

Model serving: FastAPI, BentoML, Triton Inference Server

Data pipelines: Apache Airflow, Prefect, Kubeflow Pipelines

ML training loop: forward pass, loss computation, backpropagation, optimizer

Train/val/test split discipline; cross-validation best practices

Feature engineering principles and data leakage prevention

Model drift: concept drift vs data drift; statistical detection methods

Containerization of Python ML environments (reproducibility is the core MLOps value)