CoE - Digital Engineering - ML OPS Engineer

Key Responsibilities

1. SageMaker Pipelines & Model Monitoring

• Build and maintain SageMaker pipelines for embedding generation and NER workflows.
• Extend pipelines for new workloads such as query reranking.
• Implement end-to-end pipelines across Dev, QA, and Prod.
• Develop custom monitoring (drift detection, latency, failure alerts) using SageMaker + AWS Lambda.
• Enable A/B testing and production rollout of new models.

2. GPU/CPU Performance & Cost Optimization

• Deploy and optimize GPU instances for high-throughput inference workloads.
• Benchmark and select optimal instance types for:
  • Reranking models
  • Embedding pipelines
  • Vision inference systems
• Implement Spot Instance strategies for large-scale workloads.
• Optimize:
  • Batch sizes
  • Memory allocation
  • Concurrency
• Manage inference services using Gunicorn, Boto3, and SageMaker Endpoints.

3. Search Infrastructure (Elasticsearch/Lucene)

• Optimize hybrid search (BM25 + vector search).
• Tune:
  • Index configurations
  • Sharding strategies
  • Query performance
• Collaborate on ANN/HNSW tuning (ef_construction, M).
• Balance recall, latency, and memory usage at scale.

4. Video & Image Pipeline Infrastructure

• Deploy and scale pipelines for:
  • Video shot detection (TransnetV2)
  • Image embedding generation
• Containerize and autoscale workloads.
• Handle large-scale I/O efficiently.
• Monitor latency, failures, and drift across media pipelines.

5. ML Deployment & Platform Engineering

• Define standards for ML deployment and CI/CD pipelines.
• Build deployment workflows across Dev, QA, and Prod.
• Implement automated health checks and alerting using AWS Lambda.
• Ensure consistent, scalable deployment practices.

6. Production Deployment & Cloud Operations

• Lead deployment across:
  • AWS EC2
  • AWS Lambda
  • SageMaker Endpoints
• Select optimal deployment strategies based on:
  • Latency
  • Throughput
  • Cost

Required Qualifications

• 5+ years as an ML Engineer or MLOps Engineer in production environments
• Strong experience deploying PyTorch and TensorFlow models
• Hands-on expertise with:
  • AWS SageMaker (pipelines, endpoints, monitoring)
  • AWS Lambda
  • EC2 and cloud infrastructure
• Proven experience with:
  • GPU/CPU optimization and benchmarking
  • Memory management and batch tuning
  • Spot Instance cost optimization
• Experience with Elasticsearch/Lucene for large-scale search systems
• Expertise in containerized deployments and autoscaling
• Strong understanding of feature engineering for BERT-based models
• Experience with model monitoring, evaluation, and A/B testing

Preferred Skills

• CNNs, diffusion models, and deep learning architectures
• Ranking systems (cross-encoder / bi-encoder)
• Approximate Nearest Neighbors (HNSW)
• Clustering (K-Means, DBSCAN)
• Regression, decision trees, Bayesian methods
• Experience with multimodal ML systems