Ray S

I’m a Senior Machine Learning Engineer with 6 years architecting and deploying end-to-end ML systems in production. My deep expertise spans Python ML frameworks (PyTorch, TensorFlow, Hugging Face) and production-ready services (FastAPI), with a strong focus on LLMs, RAG, and fine-tuning.

I build LLM products that behave well in the real world: RAG systems with hybrid retrieval, reranking, and source attribution; multi-agent workflows using LangGraph and CrewAI with tool calling, memory, and human-in-the-loop checkpoints; and evaluation/guardrail layers for non-deterministic systems. I’ve raised answer-quality scores by 28% while flagging drift before it reaches users, and I’m comfortable owning the path from research notebook to multi-region production endpoint.

Across AWS and GCP, I lead full MLOps lifecycle work—deployment pipelines on SageMaker/Bedrock and Vertex AI, monitoring with Prometheus/Grafana and Evidently AI, and drift detection that protects accuracy over time. I’ve also mentored 4 ML engineers and introduced MLOps standards (DVC, MLflow, Kubeflow), lifting deployment frequency 60% and cutting rollback incidents by half, while optimizing models with quantization and ONNX Runtime for faster inference.