Below the prompt

My AI work starts below the prompt. Supervised and unsupervised learning, neural networks, natural language processing, computer vision, feature engineering at petabyte scale. Then came transformers, fine-tuning, LoRA, diffusion and multimodal models, and inference optimization under cost and latency targets that a demo never has to meet.

That foundation shows up in specific work: AWS SageMaker platform comparisons and production pipelines for deep learning, speech, vision and chat AI; a GPT-based portfolio rebalancer built on the Sharpe model for a Wall Street client; a 360-day flight delay predictor built on FAA data.

None of this is prompt engineering. It is the discipline prompt engineering sits on top of.