Software Engineering student at the University of Guelph, looking for a Winter 2027 software engineering co-op. I like building things that are ambitious for their size — a full offline app, and the AI engine underneath it. Most of what I know I learned by shipping these two end to end.

A polished, fully offline dinosaur-and-space encyclopedia for kids — 100 hand-written entries, an AR Scan Sky mode that labels the real stars, planets and constellations you point your phone at, dino battles, quizzes, and Ask Nova, an AI chat that runs entirely on the device with no internet.

• An instant answer engine sits in front of the model: most questions are answered immediately from a vetted encyclopedia and a 156-topic knowledge base; only genuinely open-ended ones reach the LLM.
• Backed by an in-repo harness that runs over a billion generated questions through the exact production pipeline — each reply graded for being alive, on-topic and substantial — and passes with zero quality failures.
• Every screen is written in C# rather than XAML page markup, over one small reusable component kit, so the whole app can switch themes live.

The on-device AI inference engine I built to power DinoSpace, running Google's Gemma locally. Packaged as an Android library and bridged into .NET MAUI through a native binding.

• Streams answers token-by-token with a single-inference queue and automatic engine reload between questions, so the chat can never hang or leak state — all with no network.
• Ships a 3 GB model through Google Play as four 1 GB asset packs to stay under Play's per-pack cap, with a resumable in-app download as the fallback.

Languages: C# · Java · Kotlin · C · Python · JavaScript/TypeScript Building with: .NET MAUI · XAML · Android · React Native · Expo · Git · Visual Studio · Android Studio

📧 Karthikeya0923@gmail.com · 💼 linkedin.com/in/karthikeya0923