UIUC Electrical Engineering · ML · Embedded Systems

Fauzan A. Ishtiaq

I build systems end-to-end—hardware, firmware, and the ML that runs on it. Co-Founder at Lotus Bionics. Researcher at UIUC. GPA 3.8.

See My Work Get in Touch

Work & Research

Lotus Bionics

Co-Founder & Technical Lead

Jan 2026 – Present

Building an affordable multi-grip prosthetic hand—selected for the UIUC Cozad New Venture Challenge.

  • Architecting full-stack biosignal system: EOG + EMG analog front-end → BLE → on-device ML → motor actuation on a 3D-printed prosthetic.
  • Designing dual-sensor wearable with real-time grip classification; led technical requirements definition and competitive analysis of existing prosthetic devices.
Biosignal Processing EMG / EOG BLE On-device ML Embedded C

WaggleNet Research Group, UIUC

Researcher

Oct 2025 – Present

IoT + ML system for honeybee colony health monitoring deployed in active hives.

  • Designed custom PCB with piezoelectric accelerometer and voltage-clipping circuit driving a Raspberry Pi Pico, sealed for in-hive deployment.
  • Built signal pipeline—700 Hz LPF → FFT → spectral features—feeding ML classifiers for Varroa mite detection, bee behavior, and queen presence.
Custom PCB IoT Signal Processing FFT ML

University of Illinois Urbana-Champaign

B.S. Electrical Engineering  ·  Minor in Computer Science

Expected May 2027 GPA 3.8 / 4.0

Machine Learning · Analog & Digital Signal Processing · LLM Reasoning for Engineering · Data Structures & Algorithms · Probability & Statistics

Projects

Real-Time Audio Source Separation

Python C++ PyTorch ONNX Runtime

STFT-based pipeline that isolates speakers from mixed audio in real time. Trained a U-Net on 40+ hours of LibriMix—achieved 12.3 dB SI-SNR improvement on two-speaker mixtures. Optimized with ONNX Runtime and ring-buffer overlap-add reconstruction; sustains <30 ms latency on a single CPU core.

12.3 dBSI-SNR gain
<30 mslatency
40+ hrstraining data

TinyML Wake-Word Detection on ESP32

C TensorFlow Lite Micro ESP32

Keyword-spotting neural network running fully on-device on an ESP32. On-device mel-spectrogram computation from 16 kHz audio; int8 quantization to fit within 250 KB RAM. Achieves 93% accuracy on Google Speech Commands with <200 ms end-to-end latency.

93%accuracy
250 KBRAM footprint
<200 msinference

Skills

Languages

Python C / C++ MATLAB SQL Bash

ML & Data

PyTorch TensorFlow Lite scikit-learn NumPy / SciPy OpenCV ONNX

Hardware & Embedded

ESP32 Raspberry Pi Custom PCB Design BLE Analog Front-End

Tools

Git Linux Docker MATLAB / Simulink

Get in Touch

Open to internships in ML systems, embedded, or hardware engineering.

✉ ishtiaq.fauzan@gmail.com in linkedin.com/in/fauzan-ishtiaq </> github.com/fauzanishtiaq