Giselle McPhilliamy
Computer Science + Machine Learning @ Georgia Tech
Passionate about machine learning, software engineering and building innovative solutions. Currently pursueing a Master's in Machine Learning at Georgia Tech.
About Me
Graduation w/ friends!
Hi there, happy you are here! I am a Computer Science graduate student at Georgia Tech with a strong foundation in Deep/Machine Learning, software engineering and full-stack development. I am interested in the intersection of healthcare and AI and developing innovative technologies to improve patient outcomes.
My experience spans from curating data, training, developing and evaluating models to developing scalable applications for the web. I am always eager to learn about new technologies and work on new projects.
When I am not working I also enjoy giving it my all on the pickleball court, exploring outside, eating at the best restaurants and cooking some delicious meals too.
Experience
Graduate Teaching Assistant: Machine Learning
Quiz Development / Edstem TA
- Host weekly office hours to guide 500+ students through homework + machine learning concepts
- In charge of ensuring students receive clarification for logistical and technical question on Edstem
Machine Learning Engineering Intern
Built a patent-pending hyperspectral CNN pipeline and deployed real-time edge inference on Jetson hardware.
- Designed a spectral-spatial CNN with multi-task heads, achieving >90% classification accuracy and initiating patent filing
- Engineered preprocessing (denoising filters, normalization) and jitter augmentation to improve signal quality and generalization
- Converted PyTorch models to ONNX, optimized with TensorRT, and deployed real-time inference on Jetson Orin Nano
- Integrated Jetson with a Raspberry Pi DAQ pipeline and presented results to the CEO, senior leadership, and patent counsel
Undergraduate Researcher
Researched transformer models for cancer genomics to predict immune resistance, antigen binding, and outcomes.
- Fine-tuned DNABERT, TransVCOX, and Nucleotide Transformer on oral-cancer genomic datasets
- Identified candidate immune-resistance biomarkers via PCA, clustering, and supervised learning
- Built reproducible preprocessing and training pipelines for large-scale genomic sequences
- Developed evaluation notebooks and visualizations to interpret model predictions and derive insights
Software Engineering Intern - Rapid Development Team
Shipped a pluggable bug-triage framework and UI that streamlined fix recommendations for engineering teams.
- Architected a plugin system to analyze bug tickets and recommend likely fixes
- Implemented heuristic-based plugins to rank remediation strategies and accelerate resolution
- Wrote comprehensive tests and CI checks, enabling a smooth, low-risk deployment
- Presented a deep-dive to ~25 engineers and supported onboarding and adoption
Software & Hardware Engineering Intern
Built hardware–software tools to accelerate ex-vivo prototype testing and automated device data analysis.
- Developed and deployed a robotic test device, reducing ex-vivo test cycle time by >30%
- Automated trend analysis with MATLAB scripts, cutting analysis time by >90%
- Built a baseline logistic regression model to verify signal learnability and inform design
- Integrated Raspberry Pi + Arduino control with CNC shields and servos for precise motion control
Projects
Mobile diagnosis app for Emory Hospital that guides trainees through emergency scenarios and ranks likely diagnoses in <750 ms with a k-NN–backed API.
- Led a 6-person team through full-stack design, development, and delivery
- Built a Firebase backend with RESTful CRUD endpoints following Agile practices
- Implemented k-nearest neighbors to estimate diagnosis likelihood and suggest treatments
Fine-tuned Stable Diffusion with CLIP conditioning to generate fashion images from text, achieving an Inception Score of 8.11 and improved semantic alignment via cross-attention layers.
- Built a text-to-image pipeline using CLIP embeddings + Stable Diffusion; evaluated with IS = 8.1 ± 1.1
- Implemented cross-attention conditioning and ran ablations on embeddings and data formats
- Fine-tuned with AdamW and optimized training under A100 GPU memory constraints
Android Frogger-style arcade game built in Java with SOLID design practices, delivered over five sprints with robust tests and code reviews.
- Implemented core gameplay, collision logic, and level mechanics as an Android app
- Led a 5-person team through product design and sprint planning across five 2-week sprints
- Maintained quality with 30+ code reviews and 15 JUnit tests
End-to-end ML pipeline to flag at-risk patients using a 70k-record cardiovascular dataset, with robust preprocessing, dimensionality reduction, and supervised/unsupervised modeling.
- Cleaned and standardized data, engineered BMI, and applied PCA while preserving >90% variance
- Benchmarked Logistic Regression and Random Forest; tuned RF to >75% accuracy
- Explored K-Means and GMM with elbow/silhouette/Davies–Bouldin, analyzed feature importances and implemented DBSCAN to reduce false negatives