B.Sc. in Electronics Engineering Candidate — German University in Cairo (GUC)
Cumulative GPA: 1.27 (German Scale: Best 0.7, Worst 5.0) · Full-Tuition Academic Scholarship (Top 150 Nationwide)
Co-Head, Hardware for AI Committee — IEEE GUC Student Branch
📧 ziad.abaid@student.guc.edu.eg · 🔗 LinkedIn · 📍 Cairo, Egypt
My work bridges Digital Hardware Acceleration for AI, Hardware-Software Co-Design, Silicon Photonics, and Mixed-Signal / RF Engineering:
-
FPGA & SoC Neural Network Accelerators: Architecting parameterized pure-VHDL RTL engines for Convolutional Neural Networks on Xilinx Zynq-7000 (
XC7Z020) and Artix-7 (XC7A100T) platforms—including multiplierlessCFGLUT5lookup-network + Dadda tree datapaths (0 DSP48 slices),Q8.8fixed-point streaming pipelines with custom 144-bit single-cycle BRAM ROMs, and PetaLinux 64-bit AXI-DMA / high-speed SPI co-design. -
Silicon Photonics & Optical Interconnects: Electromagnetic modeling and S-parameter extraction of serially coupled SOI microring resonators in Ansys Lumerical (
FDTD,MODE,DGTD,INTERCONNECT) and computer-vision-driven motorized fiber-to-chip alignment (YOLOv8,$0.6^\circ$ angular accuracy) at the GUC Micro Optics Laboratory. - Embedded DSP, RF & Semiconductor Physics: Edge biomedical IIR filtering on ESP32/Raspberry Pi 5 IoMT networks, full-wave 3D EM microstrip filter synthesis in CST Microwave Studio, quantum-well laser diode TCAD optimization in Silvaco Atlas, and discrete analog circuit design.
| Project / Repository | Target Platform & Tools | Key Quantitative Results |
|---|---|---|
|
FPGA-CNN-Inference-Accelerator 5-Layer Q8.8 Fixed-Point Streaming CNN Accelerator |
Artix-7 FPGA (Nexys 4) & Raspberry Pi 5Pure VHDL RTL · AXI-Stream · 8 MHz 16-bit SPI · Python
|
• 100/100 (100%) bit-exact parity vs. Python Q8.8 golden model• ~15× fewer clock cycles ( 35k @ 100 MHz vs. 528k on Pi 5 TFLite FP32)• 0.002 ms ( ~200 cycles) post-SPI pipeline flush · 0 WNS violations
|
|
Multiplierless Zynq-7000 CNN SoC Accelerator (IEEE SSCS Design Competition — Private Repo) |
Zynq-7000 SoC (ZedBoard XC7Z020)Pure VHDL RTL · CFGLUT5 · Dadda Trees · PetaLinux AXI-DMA
|
• 0 DSP48 slices & 0 core BRAMs across 5 hardware profiles (15–65 mW @ 125 MHz)• 577 bit-exact frames verified on silicon ( 0.123–0.145 ms/frame DMA latency)• 70.08% CIFAR-10 accuracy with hybrid C++ ARM classifier & live Web GUI |
|
Silicon-Photonics-Ring-Resonators Coupled Microring Resonator PIC Modeling |
Ansys Lumerical (FDTD, MODE, INTERCONNECT) · Python / YOLO
|
• Multi-port S-matrix extraction (bus-ring, ring-ring) & C-band FSR/$Q$-factor analysis• Automated motorized stage fiber alignment ( |
|
Smart-Medical-IoMT-System Distributed IoMT Edge DSP & Telemetry Gateway |
ESP32 (C++ Dual-I2C/OneWire) & Raspberry Pi 5 (Python, TLS MQTT, SQLite) |
• On-chip • 25 FPS real-time ECG/$\text{SpO}_2$ GUI & |
|
RF-Analog-Semiconductor-Engineering RF Microstrip LPF, QW Laser TCAD & Class-AB Amp |
CST Microwave Studio · Silvaco Atlas · KiCad & Proteus | • 5.0 GHz 7th-Order Chebyshev LPF: Kuroda transforms & 3D EM optimization ( • InGaP QW/BH Laser TCAD: • 4-Stage Class-AB BJT Amplifier: |
|
Lead-Lag-Control-System-Optimization Closed-Loop Lead-Lag Synthesis & Simplex Optimization |
MATLAB & Simulink (Control System Toolbox, fminsearch) |
• 5-parameter Nelder-Mead Simplex optimization achieving |
|
The-Pianist-Audio-DSP Polyphonic Audio Synthesis & FFT Noise Cancellation |
Python (NumPy, SciPy FFT/IFFT, SoundDevice, Matplotlib) |
• Multi-octave harmonic synthesis ( |
- HDLs & Digital Hardware: VHDL (Pure RTL, Pipelined Datapaths, AXI4-Stream, Custom BRAM/LUT Primitives,
CFGLUT5, Dadda Trees, CDC), Xilinx Vivado, ZedBoard (Zynq-7000 SoC), Digilent Nexys 4 (Artix-7 FPGA), Embedded PetaLinux. - EDA, Photonics & Physics Simulation: Ansys Lumerical (
FDTD,MODE,DGTD,INTERCONNECT), CST Microwave Studio (3D EM), Silvaco Atlas (TCAD), Keysight ADS, KiCad, Proteus SPICE, MATLAB & Simulink. - Embedded & Software Engineering: C/C++, Python (
NumPy,TensorFlow/Keras,TFLite,SciPy,OpenCV,YOLOv8), 8086 Assembly, Linux/Bash, Git, SPI / I2C / UART / AXI-DMA / MQTT (TLS).
