Subhan Chaudry

Projects

Selected work

  1. 2026 · Cloud · Machine Learning · Featured

    Predicting Workload Instability in Cloud VMs

    Solo project · M.S. AI, UT Austin

    Rather than react to CPU thresholds after they're crossed, this project models VM telemetry to predict spikes before they happen, forecasting whether a cloud VM will exceed 85% CPU within 15 minutes. A Random Forest reached 99.66% accuracy, and SHAP analysis identified short-term workload momentum as the dominant signal.

    • 99.66% Accuracy
    • 0.9999 ROC AUC
    • 0.964 F1
    Editorial cover: Predicting Workload Instability in Cloud VMs — 99.66% accuracy, 0.9999 ROC AUC, 0.964 F1.

  2. 2025 · Medical Imaging · Deep Learning · Featured

    Improved Pneumothorax Classification using DCGAN Augmentation

    Solo project · M.S. AI, UT Austin

    A DCGAN synthesizes pneumothorax chest X-rays to balance a scarce SIIM dataset, and a ResNet-18 classifier is fine-tuned on the augmented data, reaching 98.6% accuracy and a perfect ROC-AUC. Grad-CAM and t-SNE confirm the model learned real clinical features.

    • 98.60% Train Accuracy
    • 1.000 ROC AUC
    Editorial cover: Improved Pneumothorax Classification using DCGAN Augmentation — 98.60% train accuracy, 1.000 ROC AUC.

  3. 2025 · NLP · Transformers

    Mitigating Answer-Length Bias in Fine-Tuned ELECTRA

    Solo project · M.S. AI, UT Austin

    A single ELECTRA-small model is fine-tuned on SNLI and SQuAD, then evaluated past its headline scores. The model degrades on long answer spans, with exact match dropping from 81% to 60% past six tokens. Targeted oversampling recovers some long-answer performance at a measurable cost to overall accuracy.

    • 90% SNLI Accuracy
    • 85.9% SQuAD F1
    • 78.1% SQuAD EM
    Editorial cover: Mitigating Answer-Length Bias in Fine-Tuned ELECTRA — 90% SNLI accuracy, 85.9% SQuAD F1, 78.1% SQuAD EM.

  4. 2023 · Accessibility · Computer Vision

    Eye Gazer — Webcam Accessibility Platform

    Accessibility · Machine Learning

    A web application that lets people with speech or mobility impairments operate communication boards using their eyes — tracking retinal movement through a standard webcam with a machine-learning gaze model, no special hardware required.

    • Webcam-only Hardware
    Editorial cover: Eye Gazer — Webcam Accessibility Platform.

  5. 2020 · Consulting · Data Analysis

    Consulting Capstone with PEAK6 Investments

    MAP Consulting · B.B.A., Michigan Ross

    A semester-long Michigan Ross MAP engagement with PEAK6 Investments to improve the firm's investment due-diligence process. Built data analysis tools in R to measure and evaluate success attributes for growth and early-stage investments, delivering a summary report to firm executives.

    Editorial cover: Consulting Capstone with PEAK6 Investments.