ANTeater — Open-Source Grocery Scanner
Why I built it
I grocery shop every day and relied on Yuka to check what's in my food, which meant paying a subscription for a lookup anyone could do with open data. ANTeater replaces it: scan a barcode and get a transparent 0–100 score, the additives driving it (with sources), and a higher-scoring product from the same category.
- Transparent score: nutrition 60 (Nutri-Score), additives 30, organic 10; any high-risk additive caps it at 49. Unit tested, method documented with sources.
- Better options: searches Open Food Facts for higher-scoring products in the same category and region.
- AI where it helps: for products not in the database, Gemini reads a photo of the ingredient list into the same scoring engine, and can explain any score in plain English.
- Spend tracking: log prices, see monthly spend and how much went to Poor or Bad items, export CSV/JSON for analysis.
- Resilient demo: installable PWA with an offline shell, on-device caching, and graceful handling of API rate limits (HTTP 429/503).
How a scan flows
Live app. Tap a saved product, or open it on your phone to scan.
Functional Tooling & Microfeatures
Every day for 80 days I ship two small, working cloud data features: one on Databricks and one on Google Cloud. Each is built on a platform feature released in the last six months, picked that morning from the official release notes. Every one has a test that runs without cloud access, lands through a checked pull request, and shows up here automatically.
■ Databricks
0 / 80■ Google Cloud
0 / 80Legacy & College Engineering Projects
NBA Clustering Project
Professional basketball has evolved past conventional positions (Point Guard through Center). This project uses unsupervised machine learning (K-Means, Principal Component Analysis, and Hierarchical Clustering) on NBA player tracking metrics to automate and enhance tactical classification, revealing hybrid role archetypes.
Slow Pitch Automated Review Calculation (SLOW-ARC)
High-speed camera tracking solutions like Hawk-Eye cost hundreds of thousands of dollars. SLOW-ARC develops an accessible, automated computer vision trajectory estimation system to determine balls and strikes in slow-pitch softball leagues, democratizing advanced officiating technology for local athletic organizations.
Sales Dashboard & Data Visualization (Tableau)
An executive business intelligence visualization suite analyzing enterprise sales performance. Features multi-level drilldowns into regional margins, customer retention curves, and revenue seasonality, transforming complex transactional records into actionable corporate insights.
Automated Schema & Primary Key Extraction Utility
Dynamic PySpark utility for programmatic column searches, metadata extraction, and primary key identification. Built for distributed teams to eliminate hardcoded SQL dependencies, cutting debugging time by over 50% with automated CI/CD schema drift alert emails.