George Luther - Data Scientist

About Me

I am a passionate Data Scientist with 3+ years of experience specializing in algorithm optimization, machine learning, and big data analytics. Proven ability to design AI-driven optimization frameworks, deep learning models, and large-scale data integrations using Python (PuLP, PyTorch), SQL (Athena), and AWS (S3, EC2). Strong expertise in statistical modeling, data validation, and Tableau-based data visualization for extracting insights from complex datasets. Currently pursuing a Master’s in Artificial Intelligence (WAM 90.67) with a focus on deep learning and predictive analytics. Passionate about scalable AI solutions, quantitative modeling, and business-driven data science.

Experience

Milton Data, North Sydney – Data Scientist

JULY 2022 – PRESENT

Developed an AI-driven optimization framework leveraging PuLP & hill-climbing algorithms to align 40M+ big data records across genre, publisher, and device dimensions using AWS Athena & S3 for IPSOS AU.
Led YouTube-CTV data integration for Google & IPSOS AU, ensuring 99.8% algorithmic consistency across multiple demographic and behavioural constraints.
Audited and validated demographic insights for IPSOS UK by assessing distributions of respondent-level data using bootstrap and other statistical methods.

Uber Technologies Ltd., Sydney – Operations Analyst

DECEMBER 2021 – JULY 2022

Automated compliance verification using Python & BeautifulSoup, reducing manual vehicle status checks for a process by 99%, saving 2-3 hours per week.
Conducted SQL-based data analysis to assess driver compliance & operational efficiency, enabling faster decision-making for business teams.
Collaborated with cross-functional teams (Legal, Compliance, and Marketplace) to optimize regulatory adherence & data-driven insights for operational KPIs.

Projects

Dementia Classification Using Deep Learning

Developed an ensemble deep learning model using MRI scans and ResNet-50 for dementia classification with limited computational resources.

Key Takeaways:

Gained hands-on experience with PyTorch preprocessing techniques.
Developed a deep learning model that achieved 59.58% accuracy across multiple dementia classes.
Worked extensively with image processing and augmentation techniques.
Handled medical imaging datasets efficiently using PyTorch image loaders.
Addressed class imbalance by implementing advanced resampling techniques.

View Kaggle Notebook Download PDF

Variational Autoencoder for Data Oversampling

Implemented VAEs to generate synthetic samples, addressing class imbalance in training data.

Key Takeaways:

Explored and visualized the latent space to understand how the VAE encodes data distributions.
Developed a Multilayer Perceptron (MLP) and a Variational Autoencoder (VAE) using PyTorch.
Generated synthetic samples to improve class balance and enhance model training.
Implemented oversampling techniques using VAE-generated data to address dataset imbalance.
Fine-tuned the VAE's loss function and regularization to improve reconstruction accuracy.

View Google Colab Notebook Download PDF

Physics Simulation of a Double Pendulum

Simulated chaotic motion of a double pendulum in MATLAB using numerical integration techniques.

Key Takeaways:

Implemented numerical integration techniques such as Euler's Method and Runge-Kutta (RK4) for simulating motion.
Modeled the chaotic behavior of a double pendulum using Lagrangian mechanics.
Developed MATLAB scripts to visualize energy conservation and phase-space trajectories.
Worked with differential equations to compute angular velocity and acceleration.
Generated animated visualizations to illustrate the sensitive dependence on initial conditions.

Download PDF

Hi, I'm George Luther

About Me

Skills

Experience

Milton Data, North Sydney – Data Scientist

Uber Technologies Ltd., Sydney – Operations Analyst

Projects

Dementia Classification Using Deep Learning

Variational Autoencoder for Data Oversampling

Physics Simulation of a Double Pendulum

Contact Me