About The Role
The role focuses on building, training, and deploying production ML systems that operate at scale - recommendation models, ranking systems, and real-time inference pipelines serving high-traffic products.
You will work within a cross-functional team of applied scientists, data engineers, and product engineers where model latency, reliability, and measurable business impact are the core success criteria, not just offline metrics.
Key Responsibilities
- Design, train, and evaluate ML models - deep learning, gradient boosting, and two-stage retrieval-ranking architectures - for production recommendation and prediction systems
- Build scalable training and feature pipelines in Python and PySpark, ensuring parity between offline training data and online serving features
- Deploy and serve models using Kubernetes, TensorFlow Serving, or TorchServe, optimizing inference latency and throughput for real-time traffic
- Instrument models with monitoring and alerting for data drift, feature skew, and performance degradation; own root-cause analysis and retraining triggers
- Design and execute A/B tests and online experiments, translating offline evaluation into statistically sound production decisions
- Collaborate with data engineers to maintain robust data pipelines in Airflow or Dagster, and with product teams to define model success criteria
- Contribute to ML platform tooling, code reviews, and internal engineering standards for reproducibility and model governance
What We Are Looking For
- 3–6 years of experience in machine learning engineering or applied ML, with multiple models shipped to production at scale
- Strong Python engineering skills; proficiency with PyTorch and/or TensorFlow plus classical ML libraries (scikit-learn, XGBoost/LightGBM)
- Hands-on experience with ML infrastructure: Kubernetes, Docker, CI/CD for models, and cloud ML stacks (AWS, GCP, or Azure)
- Solid grounding in ML fundamentals: evaluation metrics, regularization, class imbalance, and the gap between offline and online performance
- Experience with distributed training or large-scale data processing (Spark, Ray, or Dask)
- BS/MS in Computer Science, Statistics, Engineering, or equivalent practical experience
- Bonus: experience with LLM-based systems, real-time feature stores (Feast, Tecton), or MLflow/Kubeflow-based MLOps platforms