10+ years building AI systems end-to-end — from fine-tuning to production traffic. Built a LoRA fine-tuning → vLLM pipeline that saved $300K, powering fraud detection and personalization at 40k+ transactions/sec. GPU-accelerated cybersecurity ML on NVIDIA DGX (8×A100) — behavioral anomaly detection over billion-row RAPIDS datasets, phishing classification at 2× throughput, and abuse-inbox triage at 2× the prior pipeline's case volume, each threat routed to the right security team. Delivered $2.5M in annual savings via an Azure ML quality-prediction system and optimized last-mile logistics with OR-Tools. Across 10 years and six companies, I've earned the top performance rating at every one.