I build, package, and deploy reproducible, containerized machine learning workflows.

Content Refresh Prioritization — FlyRank AI ML Internship

A ranked-queue model for deciding which declining pages a content team should review first, trained on an anonymized 90-day slice of real search performance data across 32 clients. The first version scored a suspicious 100% accuracy — read as a red flag, not a win — traced to two features mathematically tied to the label. Once removed, the honest model lands at 71.4% global precision against a 62.8% base rate, and the queue gets meaningfully cleaner at the top: ~89% precision in the top 100 candidates.

Brain Tumor Classifier — live demo

A PyTorch image classification model behind a containerized Streamlit demo — reproducible to run the same way anywhere, not just demoed once and forgotten. A dedicated model-serving API (FastAPI/TorchServe) is the natural next step past this.

Engineering focus

My engineering focus is MLOps — building the infrastructure that carries a model from a notebook to something reliable enough to depend on. I'm currently working through the FlyRank AI ML internship, covering modeling, validation, and research writing against production search data.

I run my own infrastructure elsewhere — servers, domains, backup systems — and reach for it when a project actually needs that level of control. This site is a deliberately fast, static Netlify deploy instead: fewer moving parts, faster to iterate on, while the work itself stays the focus.

Connect on LinkedIn

Contact

Have a role, project, or question? Send a message below — it reaches my inbox directly.