Neural Bridge: My AI-First Portfolio with a Live Gemini-Powered Clone You Can Interview

Zelalem Gizachew — Mon, 02 Feb 2026 06:59:51 +0000

Hey DEV Community! 👋

I'm super excited to share my submission for the New Year, New You Portfolio Challenge Presented by Google AI
This challenge came at the perfect time. 2026 is all about leveling up, and I wanted my portfolio to do the same. Instead of another static page that people just skim and forget, I built a living, interactive experience: Neural Bridge (Inspired By Jarvis), where visitors can actually talk to an AI version of me in real time.

No more one-way "about me" sections! You can now interview my digital twin about my work, skills, and projects using, not even with text, But your voice while having a real-time transcription.
It's powered by Google Gemini and feels like chatting with a friend who knows my entire career story. Let's dive in!

About Me

Hi, I'm Zelalem Gizachew, a Senior Software Engineer passionate about backend systems, real-time communication, and generative AI.

I love building scalable, low-latency experiences that feel magical — whether that's streaming infrastructure, AI agents that actually understand context, or systems that bridge humans and machines seamlessly.

For 2026, my goal was simple: move beyond a traditional portfolio. I wanted a digital space that represents me dynamically, a "bridge" where people don't just read about my work… they discuss it with an AI agent that embodies my professional identity, knowledge, and style. That's how Neural Bridge was born.

My Portfolio

Live Portfolio Link: https://zelalem-portfolio-519552005901.us-central1.run.app/

Don't forget to INTERVIEW MY AI CLONE

🚀 Try it out right now!

Head to the Neural Bridge section and click the [INTERVIEW_ME] button to start a real-time voice (or text) interview with my Gemini-powered AI agent. It knows all about my skills, projects, and experiences and can answer almost any question on my behalf — from technical deep-dives to career advice.

Pro tip for the best experience: Use a wider screen (desktop/laptop) so you can see the cosmic Neural Bridge visuals without getting lost in space 😉. Headphones help for voice mode!

The aesthetic is futuristic and space-inspired — glowing connections, neural pathways, a sense of exploration — because that's how I see the intersection of real-time tech and AI: a bridge across vast possibilities.

How I Built It

The whole project was designed to showcase high-performance, low-latency AI integration in a portfolio setting. Here's the breakdown:

Tech Stack

Frontend: React + Vite — fast, modern, and perfect for the interactive "Neural Bridge" aesthetic.
Real-Time Communication: LiveKit handles all the streaming infrastructure (audio, video, data channels) so the conversation feels instant and natural.
AI Intelligence: Google Gemini (via LiveKit's Gemini Multimodal agent) — this is the brain! Gemini powers the "Almost Me" interviewer, understanding questions in real time, recalling my projects/skills, and responding conversationally.
Infrastructure: Google Cloud Run — serverless, auto-scaling home for the web frontend and the Auth server (for LiveKit auth). Deployed containerized for reliability.
Development Environment: Antigravity with — Google's AI-first dev environment. This was a game-changer!

The Antigravity Experience

I jumped into Google Antigravity (that slick agent-first IDE powered by Gemini) and it carried me smoothly through building and preparing this portfolio for deployment.
My Gemini 3 Pro trial had run out (RIP late-night experiments 🤓), so I switched to the speedy Gemini 3 Flash model, and it handled everything perfectly, keeping things fast and smart.
My favorite trick was crafting clear prompts in Gemini Web first to avoid any back-and-forth, then dropping them into Antigravity starting from an empty repo. It guided me through the whole setup: architecture, files, config, all the little details — and got the site fully ready for Cloud Run without the usual headaches. Bonus cute surprise: when I needed a quick icon, Antigravity quietly used Nano Banana to generate one for me. Nice touch! ;-)
Overall, it felt like a reliable, thoughtful teammate that made the whole process way more enjoyable.

What I'm Most Proud Of

I love the Jarvise AI Tony Stark used on Iron Man Movie, and I always wanted to build Something like that. And ever since I've learned about building websites, my goal was set for my portfolio website. I wanted to build a Jarvis like Portfolio website because It kinda make me feel like Iron Man 😁
But those animated components are expensive to built even in this AI era. I even learnt how I can manually do SVGs and I've made 1 component in almost like 2 days and it was tiresome.

But Now, with the Help of AI, I'm able to achieve this, or at least I'm closer.

And the Handshake Logic in the backend. I implemented a custom RoomAgentDispatch system that instantly dispatches my personalized Gemini agent the moment a user joins the LiveKit room. No manual setup, no waiting — seamless "human-to-AI" transition.

It feels like magic: you click INTERVIEW_ME, and "I" (My AI Clone) greet you instantly, ready to talk about anything from my backend architecture patterns to real-time AI challenges.
Who knows? In the future, I might not even need to hop on calls for initial interviews — the bridge handles it! 😄

Deployment Details

The entire portfolio is containerized and running on Google Cloud Run for effortless scaling and zero server management.

To meet the challenge requirements, I deployed with the required label: --labels dev-tutorial=devnewyear2026

Thank you for organizing this challenge.

DEV Community: Zelalem Gizachew