An interactive 3D portfolio: a hand-drawn paper world you walk through to see my projects, background and contact links.
This is the portfolio of Sree Varshan V, who goes by XENOZ online. The 3D version starts on a small island map, where a paper house lights up based on my latest GitHub activity. From there a corridor leads to four rooms: the Gallery (projects), the Studio, About Me, and Contact.
It is built with React, React Three Fiber, Three.js and GSAP, and adjusts its quality to the device it runs on.
I am an AI & ML undergraduate Computer Science student from Chennai with experience in Python, JavaScript, machine learning, and full-stack development. I have built AI-powered, embedded, and systems projects across trip planning, coding assistance, research automation, and accessibility tech, from low-latency Rust streaming pipelines to ESP32 accessibility hardware.
Pocketdesk is an open-source remote desktop (Rust, Kotlin) that streams and controls Linux (Wayland, X11) and Windows computers from an Android phone or another computer, using hardware H.264/HEVC encoding over QUIC and iroh hole punching, with no port forwarding or account. It grew into a KDE Connect-style ecosystem: clipboard, files, notifications, the phone as a webcam and second screen, and one keyboard and mouse across computers. Shared-input latency over Wi-Fi went from about 25 ms to 3 ms, and its packet parsers are fuzzed nightly in CI.
ECHOLINK is a real-time chat application built with React, Supabase, and Electron for the web and Windows desktop, featuring end-to-end encrypted messaging. It integrates an AI assistant for smart replies and sentiment analysis, plus collaboration tools including code sharing and execution, voice messages, and a shared whiteboard.
EV Qonnect is an EV trip planning and charging platform with battery-aware routing, real-time charger availability, wait-time predictions, and booking. It spans a Next.js owner portal (station management and utilization analytics) and a Kotlin Android app for drivers.
xencode is a terminal-native Rust coding agent built for free, open-weight models. It runs on Ollama, llama.cpp, a free-tier Colab GPU, or any OpenAI-compatible endpoint, and keeps your code on your machine. A sequential provider fallback chain keeps sessions alive when a model is down, and an approval-gated agentic tool loop plans, edits, tests, and fixes code, showing every diff and command before it runs. The single Rust binary spans 20 crates, with a ratatui terminal UI and Zed integration through the Agent Client Protocol.
The rest of my work is on my GitHub profile.
I am open to roles, internships and collaborations. The resume above has my full details, and you can also reach me on LinkedIn or GitHub.