Jana Trpkovska

Real small business I own — handmade candles

Lit Candle

A full storefront and admin system built around how my own candle business actually runs: products, custom scents and colors, made-to-order stock, orders, and customer communication. Built solo, from database schema to UI.

My role — Designed and built solo — database schema, business logic, storefront, admin tools, and the automated test suite.

• Public storefront plus a protected admin area covering products, taxonomy, scents, colors, and bulk editing
• Order status changes go through a fixed transition table enforced on the server, not just hidden or shown in the UI
• Bouquets are a dedicated product type with their own sizes, pricing, and box inventory — not assembled from other catalog items
• Order confirmation never depends on email succeeding: the order is committed to the database first, email sent after
• Built out an automated test suite covering core flows, alongside TypeScript typecheck, lint, and production build checks

University project — Introduction to Data Science

OpenDataScope

A research and analytics web application comparing North Macedonia's open-data ecosystem with international benchmarks, featuring interactive data visualizations, country comparisons, documented methodology, and reproducible data validation.

My role — Author and developer — research question, data collection and analysis, and the web application.

• Compares North Macedonia's portal (data.gov.mk) with six national portals, with the EU portal shown separately as an aggregator
• Interactive charts with chart/table views, plus a Compare page for side-by-side portal comparison
• Comparisons are labelled Direct, Partial or Context only, because portals count different units
• Documented methodology, source register, and a fixed, timestamped research snapshot
• Frozen local CSV data validated at build time, with SHA-256 checksums and downloadable tables

University project — interactive scene analysis + computer vision

RoadWatch

An interactive tool for laying out and analyzing accident scenes: a canvas-based editor for placing and adjusting vehicles by hand, connected to a computer-vision backend that can detect vehicles from an uploaded photo and drop them into the same scene automatically.

My role — Built both sides — the frontend scene editor and the Flask/computer-vision backend.

• Canvas scene editor: place, select, drag, rotate, resize, zoom, and pan vehicles, plus distance measurement and per-vehicle notes
• Undo/redo history, and scene save/load and image export as JSON and PNG
• A real Flask /analyze endpoint, called from the frontend with an uploaded image via fetch and FormData
• YOLOv8 detects vehicles in the photo; OpenCV (Canny edges + PCA) estimates each vehicle's orientation
• Detected vehicles are converted into the same data structure the manual editor uses, so AI output and manual edits share one scene model

A computer-vision integration and visualization tool, not a certified or legally admissible accident-reconstruction system.

My first AI-assisted project

Guess the Flag

A flag-guessing quiz with difficulty levels, hints, and a live scoreboard. An early AI-assisted project that marked the start of experimenting with faster development workflows — less polished than Lit Candle or RoadWatch, but a real, working app.

My role — Solo, built early in my AI-assisted development experience.

• Vue 3 + Vite quiz app with Easy/Medium/Hard difficulty levels and a per-round timer
• Hint system with a cooldown, plus live correct/incorrect score tracking
• Scores written to and a leaderboard streamed live from Firebase Firestore
• Player name is captured and stored locally per session — there's no real authentication behind it