MS Electrical & Computer Engineering student focused on hardware engineering — the silicon behind intelligent systems, from embedded devices to high-performance chips. I like working close to the metal, and I'm especially interested in how compilers, memory systems, and CPU/GPU architectures shape real-world performance — and how those designs are verified before tape-out.
Currently focused on design verification and embedded systems, with a growing focus on computer architecture and memory systems for AI and HPC.
- Hardware / VLSI: digital design, Verilog, design verification, embedded systems, sensor integration
- Architecture & systems: CPU/GPU architecture, memory hierarchies & caches, CUDA, OpenMP, performance analysis
- Languages: C, C++, Python, CUDA, Verilog
- AI / ML: PyTorch, applied machine learning
Real-time pressure-mapping system — Penn State Health. Built from the ground up: embedded hardware, sensor integration, and wireless visualization, working under real clinical conditions. A hands-on reminder of how small hardware decisions ripple up into system-level behavior.
| Project | What it is |
|---|---|
| vlsi-16bit-multiplier-cadence | Transistor-level 16-bit multiplier in Cadence Virtuoso (45 nm gpdk045) — CMOS gates → Booth-Wallace adder tree, Spectre-verified, Pegasus DRC |
| batched-fft-benchmark | Hand-written batched 1D FFT in C++/OpenMP & CUDA, benchmarked vs. Intel MKL and cuFFT on P100/V100 GPUs — with roofline & memory-bandwidth analysis |
| design-exploration | 18-dimensional CPU design-space exploration in SimpleScalar, optimizing for execution time and energy efficiency |
| cache-simulator | Configurable set-associative CPU cache simulator in C (variable size / associativity / block size, LRU) |
| cordic-signal-processor | CORDIC-based signal processor — complex FIR filter + Cartesian→polar rotator, built with HLS pipeline/unroll pragmas |
| five-stage-pipelined-cpu-verilog | Five-stage pipelined CPU (IF/ID/EX/MEM) in Verilog for Xilinx FPGAs |
| mapreduce-wordcount-c | Multithreaded MapReduce-style word count in C with a bounded producer–consumer buffer |
| taxi-driver-trajectory-lstm | PyTorch LSTM for GPS trajectory sequence classification (applied ML) |
- Email: arsalansaif25@gmail.com
- LinkedIn: arsalansaif15
- GitHub: @arsalansaif