At Negin Bazr Danesh, I developed computer vision pipelines for agricultural satellite imagery, fine-tuned UNet, SegFormer, and YOLOv8 models, and improved field-boundary detection and crop segmentation to 85% mIoU. I also developed super-resolution models that achieved 30 dB PSNR on evaluation data.
At Mahya Energy, I developed computer vision classification systems for livestock health monitoring on NVIDIA Jetson Nano hardware. The models achieved over 90% accuracy across primary diagnostic classification metrics in a real-world farm IoT deployment.
I built LexBook AI, an educational platform with multi-agent workflows and RAG, and Molecule Scout, a drug discovery research assistant with a retrieval-conditioned generation and docking pipeline. My work also includes an IEEE-published multimodal AI architecture for predicting movie popularity at edge nodes.

