Summary
What you’ll impact
The role focuses on building and shipping production machine learning systems, owning models end‑to‑end from data pipelines to deployment and monitoring. It sits at the intersection of software engineering and applied ML, shaping how models reach users at scale.
Responsibilities
What you'll do
- Design, train, and productionize ML models across ranking, classification, and forecasting workloads using PyTorch and scikit-learn
- Build scalable data and feature pipelines in Python and Spark, ensuring consistency between offline training and online serving environments
- Deploy and operate models on cloud infrastructure (AWS or GCP) using containerized services, Kubernetes, and CI/CD pipelines
- Implement model monitoring and alerting for data drift, latency regressions, and prediction quality degradation
- Design and run online experiments — A/B tests and shadow deployments — to validate model improvements against business metrics
- Optimize inference performance through model compression, quantization, batch tuning, or hardware-aware serving strategies
- Contribute to engineering standards through rigorous code reviews, system design docs, and shared tooling
Requirements
What you’ll bring
- 3–7 years of experience in machine learning engineering or a closely related field, with multiple models shipped to production
- Strong Python engineering skills and deep hands-on experience with PyTorch or TensorFlow
- Production experience with cloud ML platforms: AWS SageMaker, GCP Vertex AI, or equivalent self-managed serving stacks
- Solid MLOps fundamentals: experiment tracking (MLflow or W&B), model versioning, CI/CD for ML, and infrastructure-as-code
- Strong grasp of ML fundamentals: evaluation methodology, regularization, class imbalance, and distribution shift
- BS in Computer Science, Engineering, Statistics, or equivalent practical experience; MS/PhD a plus
- Bonus: experience with LLM fine-tuning and serving (vLLM, Triton), streaming data systems (Kafka, Flink), or distributed training (DeepSpeed, Ray)