At NDA, I developed and integrated LLM solutions in Python, including a text-processing pipeline using local quantized models served through vLLM. I also built retrieval over embeddings in a vector database to narrow context when the model struggled with the full reference catalog.
I built agent workflows with tool calling and reference sets to evaluate answer quality after prompt or model changes. My work also included prompt engineering, low-code workflow automation, and logging, metrics, and dashboards for containerized services.
Alongside AI/ML work, I developed Delphi (VCL) desktop modules with MS SQL Server, taking tasks from client requirements to release. For several months, I assigned tasks and priorities, tracked deadlines, and resolved blockers.
