I build things end-to-end — from ETL pipelines and real-time data engines to gamified web apps and privacy-first tools. Currently looking for a Data Science / Analytics / Software Development internship.
A code-reconstruction game: pick a language and difficulty, and the backend shuffles a real code snippet into its individual lines. Reassemble it in the correct order to win, with hints as scaffolding when you're stuck. Question bank spans 7 languages (Python, JavaScript, Java, SQL, Bash, Linux, Git) across 3 difficulty tiers. A separate ambient media layer streams synced background video + music, sourced live from YouTube.
Game design: shuffles pre-split code "blocks" per question, checks reconstructed order against a stored answer key — puzzle difficulty comes from the source data, not extra game logic.
System design: media control state lives in a polled config file, decoupling the settings panel from the game canvas; YouTube ingestion retries across a browser-cookie fallback chain to survive yt-dlp auth failures, extracting a synced mp4+mp3 pair per clip.
An async pipeline ingesting live Twitch chat from Vedal987's stream via IRC + WebSockets into MySQL, handling concurrent writes under bursty message volume. A real-time analytics engine tracks message velocity and subscriber events into a live "Hype Score," rendered on a dashboard refreshing every 5 seconds.
System design: async producer/consumer pattern to keep ingestion non-blocking under bursty message volume; batched writes to protect the DB from write amplification.
A SQL learning game where you drag query blocks (SELECT, JOIN, WHERE, GROUP BY) into place instead of typing syntax, run against a real MySQL dataset of Vedal987's Twitch chat — so every query returns actual chat messages and stats, not toy rows.
System design: query blocks compile to parameterized SQL (never raw string concat) so the "game" enforces injection-safe habits by construction.
A multiplayer card game with one shared FastAPI/MySQL game-state core, wired up to four different transport layers — REST, WebSocket, GraphQL, and gRPC — running side by side so protocol trade-offs (latency, payload size, bidirectionality) are directly comparable, not theoretical.
System design: one shared game-state core with four interchangeable transport adapters — isolates business logic from transport, so protocol trade-offs (latency, payload size, bidirectionality) are directly comparable.
An anonymity network where idle browser tabs become relay nodes — embedded inside a game so players contribute bandwidth just by playing, tracked on a server-side leaderboard. Hybrid RSA-OAEP + AES-GCM encryption, multi-hop onion routing, and proof-of-presence heartbeats, using only the browser's native WebCrypto API — no client-side crypto libraries.
System design: hybrid encryption (RSA-OAEP wraps a per-hop AES-GCM key) after hitting RSA's 190-byte payload ceiling; layered onion encryption so no single relay sees full route + payload.
An end-to-end ETL system: BeautifulSoup scrapes 132 hero profiles from a community wiki, while a PandaScore API pipeline ingests 1,000+ match records into PostgreSQL. Serves a dark-themed frontend with hero autocomplete search.
System design: idempotent ETL writes (safe to re-run without duplicating match data) and parameterized queries throughout to eliminate injection risk.
A privacy-first desktop tool that parses local Google Takeout JSON/HTML exports entirely offline, surfacing top channels, keyword frequency, and viewing habits — no data ever leaves the machine.
System design: fully offline processing boundary — no network calls in the analysis path — so sensitive personal history data never leaves local disk.