• Developed a real-time FPGA-based object detection system that achieved 110 FPS on event data by designing a customnetwork and optimizing inference through threading, leading to 81% faster inference than conventional methods.• Evaluated and implemented different machine-learning models for identifying operational anomalies in live motor systems.• Implemented and reviewed Spiking Neural Networks (SNNs) for event-based vision applications.

• Developed a 96% accurate FPGA-based real-time passenger counting system with head tracking, including dataset preparation.• Reduced inference time of deep learning models by up to 90% through pruning, quantization, and knowledge distillation.• Implemented and trained YOLO object detection family from scratch (YOLOv2, YOLOv3, YOLOv4, and YOLOv6) on PyTorch.

• Developed a Nepali Automatic Speech Recognition system utilizing Whisper, capable of recognizing multiple Nepali accents.• Collected, cleaned, and compiled a comprehensive dataset for the project. Achieved a Word Error Rate below 30.