Jamie Cook

Owned the design, deployment, evaluation, and production monitoring of a large-scale LLM assistant using agentic workflows. Built MCP-enabled tool orchestration with retrieval pipelines for intent routing, semantic search, and response generation. Developed an end-to-end RAG query understanding platform and a multimodal image-to-product retrieval system using CLIP. Optimized semantic search inference with quantization and ONNX and benchmarked retrieval quality against Elasticsearch.

Developed and deployed production ML systems for ETA prediction and intelligent routing at scale. Built recommendation and retrieval components using vector search and semantic similarity methods. Implemented monitoring and evaluation services for model quality, drift tracking, and automated retraining workflows. Improved transformer inference performance via ONNX optimization and supported OCR fine-tuning for identity verification processes.

Fine-tuned and deployed NLP models for multilingual translation workflows and built reusable inference infrastructure for future projects. Developed a BERT-based conversational assistant for stock inquiry use cases. Created deep learning models for edge AI inference and optimized deployments using TensorFlow Lite. Implemented a CRF-based NER system for address matching and built sentiment analysis pipelines for high-volume social data.