Issam Alzouby

Medical AI • Digital Twins • Motion Modeling
profile

Medical AI engineer focused on real-time digital twins, motion modeling, and GPU-accelerated systems. I deploy scalable PyTorch/CUDA pipelines with FastAPI backends and modern web frontends, and mentor students on applied ML and Medical-AI workflows.

Experience

AI Research AssistantUNCC AI4Health Center | Jun 2023 – Present
  • Led deployment & optimization of motion-modeling AI pipelines, cutting inference time by 60%+
  • Built a scalable CUDA system with custom endpoints for commercial web app integration
  • Trained transformer-based text-to-motion models for realistic human movement generation
Graduate AI Research MentorStanford University | May 2025 – Jul 2025
  • Mentored a cohort of 10 high-school students in Medical AI (Stanford HAI AI4ALL)
  • Taught advanced linear algebra, ML fundamentals, and deployment of Medical-AI systems
  • Guided zero-shot medical image classifiers using BiomedCLIP & SmolVLM
AI Research EngineerDuke Heart Center & UNC | May 2024 – May 2025
  • Best Oral Presentation in Cardiac Surgery (62nd Eastern Cardiothoracic Surgical Society)
  • Designed an algorithm projecting +94% organ-donation success and reduced ICU workload
  • Co-authored 4 papers on AI-driven clinical decision-making with custom software & models
GPU Infrastructure Support TechiRepairCLT | Jan 2020 – Dec 2022
  • Repaired ASIC hashing boards and NVIDIA GPUs (RTX 30/40 series)
  • Maintained GPU clusters in harsh environments to maximize uptime

Projects

7 Time Hackathon WinnerAug 2022 – Present
  • NC NASA Hackathon, HackNC, NC State Hackathon, Truist Immersive Experience, UNCC AIR
  • 1st place, 2024 NC State Competitive Programming Competition
  • Built AI voice-translation & bias-reduction apps; 1st place in 5 major events
Home Lab — AI Fine-Tuning ServerJan 2024 – Present
  • Deployed Dell PowerEdge servers & 40TB NAS over 10GbE for low-latency workflows
  • Experimented with quantization & fine-tuning for model compression
  • Fine-tuned ResNet50 on chest X-rays for pneumonia detection (90% accuracy)
Text & Audio → Motion Model DeploymentJan 2025 – May 2025
  • Reduced T2M inference from 90s → 7s; audio preproc 4min → 30s via CUDA optimizations
  • Built FastAPI endpoints + Next.js UI for real-time T2M/A2M generation
  • Motion retargeting to TADA avatars with Three.js for in-browser 3D animation
Motion Encoder ResearchMay 2025 – Jul 2025
  • Implemented a VQ-VAE for discrete motion embeddings for transformer-based generation
  • Trained an autoencoder with low reconstruction errors (L2: 0.0159, L1: 0.0640)
  • Boosted performance with codebook resets & EMA stabilization

Technical Skills

Languages

Python (Adv), HTML/CSS (Adv), JavaScript (Int), SQL (Int), C++ (Beginner)

AI & ML

PyTorch, Transformers, Fine-Tuning, Model Architecture, CUDA Optimization, Data Processing, API Deployment

Cloud & Deployment

AWS (SAA-C03), GCP, OpenAI API, FastAPI, Next.js, Git, Hugging Face

Education

M.S. Artificial Intelligence — UNC CharlotteExpected May 2027
B.S. Computer Science — UNC CharlotteMay 2025

Awards

AI Travel Grant
Undergraduate Research Assistant
2025 Truist Student Leader
AEOP Summer Leader

Languages

  • English (Native)
  • Arabic (Native)

Interests

  • AI for Healthcare
  • Motion Simulation
  • Mentorship & Outreach
  • Hardware Tinkering
github.com/ialzoubylinkedin.com/in/Alzouby@ialzouby