At Retrocausal, I led development and deployment of production computer vision and image-based LLM systems using YOLO v8/v11, improving real-time performance and accuracy.
I also architected an automated model-training and MLOps pipeline from scratch with AWS SQS, DVC, Git, and Slack webhooks for scalable, version-controlled, monitored model iteration.
Previously, I developed computer vision solutions at Retrocausal and built chatbot and NLP systems using the BERT Transformer architecture at SMR-Tech. That work also included ML-based logistics and scheduling automation and an Employee Activity Detection project.
My project work includes a terrorism modeling and prediction research project using ensemble methods and multilabel-multioutput techniques, deployed through a Flask web portal, and exploratory analysis of a car-image dataset. I'm expanding into production LLM tools including RAG, LangChain, FastAPI, and vector databases.
