Omid Aghdaei

Played key role in developing & optimizing state-of-the-art models for audio-to-text conversion. Collaborated on project for converting telephone speech to text, leading to best-performing model in Iran. Contributed to Speaker Diarization projects, successfully identifying & separating speakers in audio recordings. Leveraged tools like VAD to improve speech segmentation. Utilized model compression techniques like ONNX, Distillation, and Quantization to optimize size, performance, and deployment efficiency.

Developed deep RL system for drone navigation, achieving 2% accuracy improvement and 18% faster inference. Utilized MiDaS algorithm for obstacle detection in diverse weather conditions. Trained ResNet-8 on real collision datasets for precise decision-making. Compressed networks with ONNX, resulting in 25% faster inference with minimal accuracy decrease. Developed specialized model with YOLO v5 for small obstacle detection, enhancing drone's navigation capabilities in complex environments.

Thesis title: Quadcopter Autonomous Navigation using Deep Reinforcement Learning Grade: 3.61/4.00 GPA

Thesis title: The relation of Internet of Things in smart city and smart car Grade: 3.31/4.00 GPA