At Network Science, I build enterprise AI systems including MySQL and MSSQL MCP servers, Graph RAG for relationship-aware retrieval, and a five-stage NL-to-SQL pipeline that blocks hallucinated or unsafe SQL before execution.
I also create agentic workflows with CrewAI and n8n for contract risk review, extracting vendor commitments and checking them against internal legal policy and liability benchmarks.
Previously at AgriSavant, I selected and fine-tuned YOLOv5 to 0.87 F1 and 0.90 mAP for pest and disease detection, then deployed computer vision services using YOLO, DINOv2, CLIP, FastAPI, AWS, and SageMaker. I’ve also improved waste-classification performance through GAN-generated imagery and transfer learning.
