About This Project

Jules Udahemuka

AI Research - ML, Safety & SciML

Twitter LinkedIn

About Me

I’m Jules, an AI researcher passionate about tackling complex challenges in Machine Learning, Scientific Machine Learning (SciML), climate science, and safety. Currently, I’m toward the end of Master’s in Artificial Intelligence at Carnegie Mellon University.

My journey in AI has been fueled by a passion for applying advanced machine learning techniques to real-world problems. Whether it’s leveraging Python and SQL to build predictive models or creating interactive dashboards in Power BI and Tableau, I enjoy translating raw data into actionable insights that drive meaningful change.

I’m a firm believer in the power of open science and collaboration. By sharing ideas and working collectively, I believe we can accelerate breakthroughs in AI and create solutions that benefit both people and the planet.

Interests

  • Machine Learning; SciML, and Safety
  • Climate Science, and Weather Prediction
  • Foundation Research
  • Open Science

Education

  • 🎓 MSc in Engineering AI, May 2017 - Sep 2021
    Carnegie Mellon University
  • 🎓 Environmental Engineering, Sep 2015 - Apr 2017
    University of Rwanda

Skills

  • Python Python
  • TensorFlow TensorFlow
  • PyTorch PyTorch
  • OpenCV OpenCV
  • Git Git
  • Linux Transformers
  • Machine Learning Machine Learning
  • Deep Learning Deep Learning
  • SQL SQL
  • Power BI Power BI
  • dbt dbt
  • Tableau Tableau
  • Research Research

Projects

Drones Computer Vision

Developed an open-source project for analyzing drone footage and applying computer vision techniques like YOLO and Faster R-CNN for real-time object detection and tracking.

Code Data

Digital Twin for Climate Change

Created a digital twin framework using AI and simulation to model and visualize the potential impacts of climate change on the African continent.

Code Data

Securing USSD Systems with Machine Learning

Implemented unsupervised anomaly detection algorithms and supervised fraud prediction models to enhance the security of USSD mobile money systems in Africa.

Code Data

Medical Imaging Hackathon Winner ($3000 in Prizes)

At Nucleate Pittsburgh BioHack, I developed Monte Carlo dropout simulations for uncertainty estimation in medical imaging models and integrated large language models (LLMs) to generate evidence-based medical suggestions based on patient history and model output probabilities.

Code Data

Neural ODEs for Regional Weather and Climate Prediction (Ongoing Research)

eveloping a novel approach for downscaling global climate predictions using Neural ODEs. Leveraged domain-specific datasets and scalable architectures to enhance predictive accuracy for developing regions..

Code

Image Classification & Verification Using Improved ResNets(with SE, Attention Layer)

mplemented ResNet paper and improved it with SE and Transform layers for image classification and face verification, achieving a classification accuracy of 90%+ in both tasks.

Code Data

Now Reading

Foundations and Trends in Multimodal Machine Learning: Principles, Challenges, and Open Questions

Paul Pu Liang, Amir Zadeh, Louis-Philippe Morency. 2022.

Paper Talk

Speech and Language Processing

Dan Jurafsky and James H. Martin. 2024.

Attention Is All You Need

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, Illia Polosukhin. 2017.

Paper Talk