Muhammad Umer Abbasi

• Developed and trained an autoencoder on an unlabeled image dataset to learn spatial features and then applied transfer learning to build an image classifier using a labeled subset of the same dataset • Fine-tuned a ResNet-152 model for 20-class image classification, using a lower learning rate for ResNet-152 layers and a higher learning rate for the custom classification head • Implemented and trained a U-Net-based 10-class image segmentation model • Developed and trained LSTM and BERT-based models

• Designed and developed schematic diagrams, PCBs, Embedded C/C++ and Verilog codes for electronic products, selected and purchased electronic components, and liaisoned with embedded software teams, QA teams, and client teams to effectively manage the project • Developed a test bench using DE-10 Nano board (Cortex A9 + Cyclone V) to acquire and analyze signals from electronic modules under test • Secured approval for a project to develop a deep learning tool for detection of manufacturing defects on PCBs

Relevant Courses: Project Lab Human Activity Understanding | Introduction to Deep Learning | System Design for IoT | FPGA Information Processing Systems | Data Networking | System-on-Chip Technologies