An AI resume & portfolio builder — fill a form, get a polished site and a one-click PDF. Real domain, real users.
●Dhairya Kumar — Full-stack · AI · Android
I ship products, not just repos.
First-year CSE student, allergic to tutorial hell. My work lives on real domains, serves real users, and occasionally beeps at fires. Current main quest: AXiS — an AI layer over everything I own.
Shipped, not shelved.
Six builds that escaped localhost — two live on real infra rn. Click any card for the lore.
A civic platform where AI reads your report — broken streetlight, overflowing bin — and routes it to the right municipal desk.
Real-time object detection and tracking — YOLOv8 + Deep SORT streaming into a live dashboard with a 3D view.
An AI translation suite — auto language detection, Gemini-powered translation, and native-accent text-to-speech.
Arduino + IR flame sensors, ~95% detection accuracy in testing, with a buzzer + LED alert chain.
Artificial eXperience & Intelligent System — a desktop AI in a floating Dynamic Island. Voice in, action out. Phase 1's alive on my desktop rn. Build log below ↓
AXiS — an AI OS for my whole machine.
Artificial eXperience & Intelligent System. A desktop AI that lives in a floating Dynamic Island pill — say the wake word, it listens, thinks with whichever LLM brain is awake (local Ollama or Claude), talks back, and actually does things — down to organizing the folder you're staring at.
Not a chatbot — a modular platform: voice, memory, desktop automation, a document brain, a coding copilot, computer vision, and IoT hooks. One brain across everything I own. Phase 1 is alive on my desktop rn — not a mockup. The log:
- core engine — FastAPI + one WebSocket event bus for everything
- voice intelligence — wake word → Whisper → agent → Piper TTS (mutes itself so it can't wake itself. learned that the fun way)
- hot-swap brains — NVIDIA / Ollama / Claude, switchable mid-conversation, no restart
- desktop control + memory — open apps, web search, SQLite long-term memory, folder automation
- dynamic island UI — frameless, click-through, always-on-top Electron pill, pure-CSS morph
- document brain — RAG over your files with ChromaDB / Qdrant
- AI coding copilot — pair-programming inside the island
- vision + device hub — OpenCV eyes, ESP32 / Arduino / MQTT for IoT & robotics
hover it. that's the whole app — a pill that floats over everything and expands when you talk to it.
follow the build ↗Learning in public, shipping in production.
The pattern: learn a thing, ship it before the motivation dies. Learning web dev became LaunchFolio — live product, real users, real 2am bug reports. A civic-tech idea became Community Hero, running on Google Cloud. One afternoon with an Arduino became a fire detector with ~95% accuracy and zero chill.
Main quest right now: AXiS — an AI operating layer with persistent memory, vision, and voice. One brain across every device I own. Everything else is a side quest, including Jetpack Compose (convincing Android that a web dev can be trusted).
B.Tech CSE @ NIET · Class of 2029
Greater Noida, India
→ Android Basics with Compose — Google, '26
→ Data Analytics simulation — Deloitte x Forage, '26
→ Shortlisted — Google Gemini Student Ambassador, '26
→ Innovate X Hackathon — NIET, '26
Got something worth building?
Internship, collab, or a weird idea that deserves to exist — say less. My inbox replies faster than my CI pipeline.
Or, beat the blob
You 0 · Blob 0 · Draws 0Tic-tac-toe. You're ✕, the blob's ○ — and it gloats when it wins.
Your move — you're ✕.