Building software that solves real‑world problems.

Opening a business in the wrong location is one of the most expensive mistakes an entrepreneur can make, and the data needed to avoid it (census demographics, competitor density, transit access) is scattered, raw, and unreadable to non-analysts.

A full-stack feasibility platform that fuses Statistics Canada 2021 Census data with live OpenStreetMap competitor, transit, and commercial-activity feeds, runs them through three Random Forest models, and streams results to eight monitoring modules over a WebSocket message bus.

A risk classifier at 88.14% accuracy, a revenue regressor at R² = 0.93, and a feasibility regressor at R² = 0.99, packaged behind an interface a non-technical founder can use, with an LLM chat for scenario questions.

The Kepler archive holds thousands of candidate signals, but a large share are sensor artifacts and eclipsing binaries masquerading as planets. Separating real exoplanets from noise by hand doesn't scale.

A complete pipeline over the NASA Exoplanet Archive: domain-driven physical auditing to remove artifacts before training, a head-to-head Random Forest vs. MLP comparison tuned with GridSearchCV, MLflow tracking on every run, and the champion model deployed on AWS EC2 behind a Streamlit frontend.

94.91% classification accuracy with 95% precision and 94% recall from the Random Forest champion, beating the MLP baseline by 3.5 points, live at eth0s.online.

Drug labels are written for everyone, which means they're actionable for no one. A warning that matters enormously to someone with a heart condition is buried in the same wall of text shown to everyone else.

A full-stack companion with a sequential dual-API pipeline: official FDA label data (warnings, adverse reactions, interactions) fetched from OpenFDA, then combined with the user's full health context in a structured Gemini 1.5 Flash prompt that enforces strict JSON output across five analysis dimensions.

Personalized, structured safety guidance, organized by interaction warnings, condition conflicts, allergy flags, usage guidance, and questions for your doctor, instead of a generic label dump.

When multiple drones approach from multiple directions, the hard problem isn't detection, it's prioritization. An operator needs to know which cluster of threats matters most, right now.

The core intelligence layer: DBSCAN spatial clustering over live threat feeds to group and rank threats by severity and proximity, served by a FastAPI backend with real-time scoring and a React/TypeScript dashboard for live tactical visualization.

A working, deployed system within the 48-hour window, live threat ingestion, clustering, ranking, and visualization on Vercel.