Ali Nawaf

Machine Learning Research Assistant

Houston Methodist Hospital

Aug 2023 – Present

• Achieved >90% accuracy in cardiovascular risk prediction by developing and training a ResNet-based CNN model on medical CT scan images.
• Engineered a signal processing pipeline to digitize raw ECG signals from 10,000+ patient records, enhancing signal fidelity for downstream ML analysis.
• Conducted a comparative analysis of SOTA deep learning architectures (e.g., ResNet, VGG) for cardiovascular risk, benchmarking accuracy and computational cost.
• Ensured 100% compliance with patient data privacy standards (HIPAA) by developing data anonymization scripts during the ECG signal processing pipeline.

Researcher

UWC Science Collective

July 2025 – Present

• Implemented and optimized variational autoencoders (VAEs) in Python using PyTorch, contributing to research in a Cornell University PhD-led lab.
• Applied advanced mathematical concepts to design and test VAEs and diffusion models, enhancing data generation.

Research Assistant

Case Western Reserve University

July 2025 – Present

• Researching privacy-preserving frameworks leveraging PCA, differential privacy, and federated learning for scalable use in agriculture and bioinformatics.

Software Engineer

Eaton

Jan 2025 – May 2025

• Engineered and deployed a fully automated data pipeline on Microsoft Azure, processing over 100,000 Salesforce records monthly and cutting data processing latency by 35%.
• Led the team to present project results, metrics, and documentation to supervisors and non-technical leadership, resulting in the adoption of new data-driven dashboards.
• Implemented role-based access control (RBAC), strengthening access control and reducing security risk.

Machine Learning Intern

Heads-up Hockey

Dec 2024 – Aug 2025

• Implemented an interactive Swift-based iOS application, integrating a Core-ML model for efficient on-device inference which improved user retention by 63%.
• Optimized the Core-ML model for on-device performance by applying 16-bit float quantization, reducing model size by 50% and improving battery-life efficiency.
• Leveraged the Core-ML framework to accelerate the model on Apple's SOC hardware (Neural Engine), achieving real-time inference.
• Automated a computer vision data annotation pipeline using Kubernetes, reducing manual data preparation time by 70% and accelerating model training cycles.

IT Intern

EarthLink ISP

May 2024 – Aug 2024

• Built a Java-based monitoring system for real-time network health tracking, improving fault detection by 50%.
• Automated support ticket classification with Python, reducing manual workload and response time by 60%.

Machine Learning Research Assistant

Houston Methodist Hospital

Aug 2025 – Present

• Independently researched and developed a novel hypothesis for cardiovascular risk prediction, applying a pioneer ML model to test the theory against a real-world CT scan dataset, culminating in 90% predictive accuracy.
• Engineered a signal processing pipeline to digitize raw ECG images, implementing open-source models and post-processing algorithms to extract meaningful signals and enhance fidelity for downstream ML analysis.

Research Assistant

Case Western Reserve University

Aug 2025 – Present

• Collaborating with Prof. Erman Ayday on a $1.2M NIH-funded project developing a privacy-preserving sandbox for federated genomic data sharing and ML-based GWAS analysis.
• Implemented methods for sample relatedness and population stratification to enhance reliability for genetic association models.

Medical Smart Labeler

PyTorch, Python, API, Docker, YOLOv10

• Reduced manual image annotation time by 60% by developing a Python tool that integrated with the Label-Studio API to pre-label images using a fine-tuned YOLOv10 model.
• Analyzed and pruned redundant layers from the YOLOv10 model, reducing computational complexity and inference latency by 30% for the real-time pre-labeling tool.

ElectroVector App

Swift, Python, API, Signal Processing

• Empowered medical staff with faster diagnostic insights by developing an iOS app that transforms raw ECG signals into clinical Vectorcardiograms, automatically extracting 5+ key cardiac risk metrics.

Generative ECG Models

PyTorch, VAEs, Signal Processing

• Investigated generative modeling techniques by implementing a Variational Autoencoder (VAE) in PyTorch to learn latent representations of ECG waveforms for synthetic data augmentation.

Medical Smart Labeler (SAM2 Version)

PyTorch, Python, API, Git, GitHub, Docker, Networking

• Reduced manual image annotation time by 60% by developing a Python-based tool that integrated with the Label-Studio API to pre-label images using a SAM2 model.

Variational Autoencoder (VAE) Implementation

Python, PyTorch, TensorFlow, NumPy, MNIST, Matplotlib

• Designed and implemented a generative deep learning model to reconstruct images and generate new synthetic data samples from a learned latent space.
• Visualized the learned latent walk in an interactive user interface.

Network Router Simulator

C, Binary I/O, Parsing, TCP, UDP, Data Serialization

• Developed a flow analyzer (hash-based NetFlow, TCP RTT), a router simulator, and an IPv4 binary trace parser in C to quantitatively analyze complex network packet data in short time.