At Big Rattle Technologies, I developed computer-vision models for retail shelf monitoring, curating and versioning shelf images and creating SKU classes. I evaluated detection performance and used bounding-box geometry and OCR validation to address product confusion.
I implemented the platform on GCP Cloud Run with Docker, Terraform, and Bitbucket CI/CD, reducing response time by 50% through Python multiprocessing. I also built an AWS Lambda pipeline to ingest sanctions data from five sources and use Mistral 7B RAG to parse PDFs into structured data.
