I build machine learning systems that turn complex data into useful decisions. My work connects model development, retrieval, APIs, and deployment, with a focus on reliability and practical impact.
Retrieval systems, applied machine learning, and the software that makes them useful.
01
RepoMentor
A local RAG assistant that helps developers understand unfamiliar codebases through natural-language questions. It combines semantic and keyword search, cites source code, and acknowledges when the available evidence is insufficient.
PythonRAGHybrid retrievalLocal LLMs
02
Production ML at Radial
Developed CNN/LSTM models for signal-quality classification and integrated inference into a clinician-facing platform. Worked with clinician feedback to move from whole-record labels to segment-level flags that better supported review.
PyTorchDjango RESTPostgreSQLDockerNginx
03
CanWeTrustReFAIR
Built a text-classification pipeline for fairness-related software requirements. Compared a TF-IDF baseline against XGBoost with BERT token features using 10-fold cross-validation on 12,401 synthetic user stories across five domains.
NLPBERTXGBoostModel evaluation
02 / Experience
Research depth. Engineering focus.
Radial
Machine Learning & Software Engineering
Worked across data preparation, model development, backend integration, and deployment. Built with PyTorch, scikit-learn, Django REST Framework, PostgreSQL, Docker, and Nginx, translating feedback from end users into changes to the product.
York University
ML Research · MSc in Computer Science
Combined mathematical modelling, simulation, information theory, and machine learning to study complex systems. Developed a random forest surrogate using 2,100 simulation samples and conducted analysis across 14 model variants.
Graduate teaching assistant for Python, object-oriented programming, and software design.
My MSc research explored how systems distinguish useful signals. That background shapes how I think about representation, uncertainty, and evaluation in machine learning. My first-author research was accepted at IEEE GLOBECOM, and my thesis was nominated for a Best Thesis Award.
04 / Let's connect
Building useful AI takes more than a good model.
I’m interested in AI/ML engineering and MLOps roles where I can work across models, software, and deployment. If that sounds like your team, I’d be happy to talk.