At Digigo, I lead the AI function and shipped a production multilingual toxic-comment classifier using a local embedding model and ChromaDB. I chose this approach to handle code-switching and non-English content without per-language retraining.
I also built a Spring Boot and MongoDB backend that runs AI moderation inline with REST APIs, removing the need for an asynchronous post-publish review queue. I’m developing an NSFW content classifier as the next moderation layer.
On Baku Sentinel, I built a flood-risk forecasting system with a leakage-free DuckDB pipeline and calibrated XGBoost model, then shipped a FastAPI inference API and dashboard. My flood-segmentation project won first place at the IT SkillSprint Capstone Hackathon.
