At Little John, I built RAG and evaluation harnesses to measure agent accuracy across providers and support data-driven model selection. I also owned the architecture of configurable AI agents, from design through Docker deployment and horizontal scaling with Ray.
At Little John, I fine-tuned LLMs and used structured validation frameworks to achieve 99% accuracy in document classification, table extraction, restructuring, and automated form filling. I integrated OpenAI, Mistral, Groq, and Qwen/Kimi to select models for different tasks.
As a freelancer, I designed and fine-tuned YOLO-based object detection models for image recognition and tracking, and deployed AI models for audio-to-text and text-to-text transformations. At Groupe Karavel-Promovacances, I developed a supervised ranking model for Google search pages and analyzed large-scale data to inform content strategy and marketing campaigns.
