David Moser

Built and operated a full-stack AI SaaS product end to end. Designed and deployed a custom LoRA fine-tuning pipeline for personalized image generation on serverless GPU infrastructure (Modal, RunPod). Took the system from research through hyperparameter optimization and iterative experimentation across training and inference to production. Implemented Python backend, TypeScript/React frontend, Stripe payments, and Cloudflare R2 storage. Launched with paying customers.

Designed and shipped a LangGraph-based LLM RAG system (Python, FastAPI, Azure) serving 3,000+ enterprise users worldwide. Delivered a 30% increase in operational efficiency and $30M in annual cost savings. Built a multimodal document ingestion pipeline using Azure Document Intelligence (layout parsing, bounding box detection, textual extraction) to process complex unstructured documents at scale and reduce downstream hallucinations. Developed a React interface for real-time document navigation, AI output verification, and citation tracking, reducing response verification time by 60%. Optimized retrieval for high recall (90%+ across RAGAS) with query expansion, term standardization, and targeted filtering. Engineered GDPR compliance via NER-driven PII/out-of-scope validation and built an Azure Functions verification pipeline enabling fact-checking through a React-PDF viewer.