At Digigo / TERZ, I develop wardrobe intelligence pipelines with RF-DETR and RTMDet for garment detection, instance segmentation, and structured outputs used by downstream product features. I also compare model quality, latency, GPU memory use, and real-image failure cases.
At Clickusta, I contribute to an AI assistant for Instagram businesses that combines catalogue-grounded RAG, visual product retrieval, multilingual responses, and human handoff workflows.
I originated the core concept for AOE-AZ, a research initiative for Azerbaijani open-ended answer assessment and personalised learning. The initiative secured a 30,000 AZN grant in 2026, and I collaborate with Prof. Shahnaz Shahbazova on research development and planning.
In my projects, I built a wildfire analysis system combining U-Net satellite segmentation, Random Forest risk analysis, NASA FIRMS data, and a Streamlit/Folium dashboard. I’ve also worked on depth-based robotic mapping and handwritten text recognition with CNN–BiLSTM–CTC.
