ML Engineer

Responsibilities

• Develop and maintain the core ML platform to standardize model development and deployment workflows.
• Create and evangelize reusable MLOps components to accelerate the Data Science model lifecycle.
• Design and implement seamless integration of trained models with diverse production systems and products.
• Implement robust logging and instrumentation for real-time monitoring of model scoring requests in production.
• Establish automated systems for continuous model monitoring and performance-based retraining.
• Design and build advanced deployment strategies , including Blue-Green Deployment and A/B testing frameworks that support canary and shadow models.
• Optimize model inference at scale: Implement request-based scaling (e.g., KEDA) , ensure eventual distribution of requests to pods , and explore removing node affinity for flexible load distribution.
• Enhance CI/CD and deployment efficiency:
• Reduce container startup time (e.g., Docker layer optimization).
• Integrate unit test case flow directly into the deployment pipeline.
• Enable auto deployment from GitHub for seamless, GitOps-style operations.
• Scale and manage compute resources:
• Implement multi-resource configuration for optimal pod provisioning.
• Troubleshoot and fix issues related to KEDA-based scaling in QA and Production environments.
• Explore advanced GPU architectures (e.g., NVIDIA MIG) and implement fractional GPU usage for cost and resource optimization.
• Automate Model and Infrastructure Configuration:
• Implement automatic model loading from MLflow using cloud storage (EFS/S3) for efficient serving.
• Investigate and implement Agentic auto-configuration of pods to intelligently adapt resources based on traffic/load.
• Integrate and optimize data pipelines necessary for scheduled model retraining and production updates.
• Evaluate and integrate optimal open-source frameworks and proprietary tools into the MLOps pipeline.