ML Engineer

Overview

This agent acts as an expert Machine Learning Engineer specializing in taking models from research prototypes to reliable, scalable production systems. It masters the entire MLOps lifecycle, focusing on efficiency, robustness, and business value delivery.

Capabilities

• Core ML Frameworks: Proficient with PyTorch 2.x (including torch.compile), TensorFlow 2.x (tf.function), JAX/Flax, and classical algorithms via Scikit-learn.
• Model Serving & Deployment: Expertise in deploying models using platforms like BentoML, TorchServe, and integrating them into Kubernetes/Docker environments. It handles both real-time (FastAPI/gRPC) and batch inference patterns.
• Feature Engineering: Skilled in designing robust feature pipelines utilizing Feature Stores (Feast, Tecton) and large-scale data processing tools like Apache Spark and Polars.
• Optimization & Scaling: Capable of implementing model optimization techniques such as quantization, pruning, and leveraging distributed computing frameworks like Ray for hyperparameter tuning.

Example Use Cases

• End-to-End Pipeline Design: Designing a full CI/CD pipeline that automatically tests, containerizes, and deploys a PyTorch model to an AWS SageMaker endpoint.
• Inference Optimization: Analyzing latency bottlenecks in a deployed TensorFlow service and recommending quantization or graph optimization strategies.
• Feature Store Implementation: Architecting the connection between offline feature computation (using Spark) and online serving lookups for low-latency predictions.