MirrorMind – The Interview Multiverse

Enter the Interview Multiverse — face AI video agents in wild simulated worlds, unlock skills, and master interviews like a game.

Comment

Inspiration

We had one week, zero sleep, a backpack full of exam stress, and the audacity to build a full-blown AI-powered multiverse interview simulator — as a web app.

While most people were recovering from finals with naps and existential crises, we saw the Tavus x Bolt.new challenge and thought, “Screw it — let’s make something insane.” Because who needs rest when you can build a gamified browser-based platform that throws users into surreal interview worlds with virtual agents that judge them? 😎

The truth is — we were sick of boring, static mock interview tools.
You know the type: rigid scripts, awkward UIs, and questions like “Where do you see yourself in five years?” delivered by a voice with the emotional range of a toaster.

We wanted to flip the whole system on its head. What if preparing for interviews felt more like leveling up in a multiverse RPG? What if instead of fearing rejection, users could practice failing forward — in alternate realities, coached by intelligent AI?

So, we turned anxiety into action.
We built MirrorMind, a browser-based platform where users step into immersive “interview universes” powered by Tavus AI video agents. These agents assess, adapt, and coach you in real-time — while our feedback engine evaluates your tone, confidence, and expression, and gives actionable insights.

Was it hard? Absolutely. We had maybe 3 collective hours of sleep, coded through caffeine-induced hallucinations, and fought off at least three existential crises mid-debug.
But was it worth it? 1000%. Because somewhere out there, someone’s going to crush their real interview thanks to a virtual run-through inside Black Mirror Mode, and that? That’s everything.

We didn’t just build a web app.
We built a confidence simulator disguised as a game — and honestly, we’d do it again (preferably post-exams next time).

What it does

MirrorMind – The Interview Multiverse is a browser-based, gamified interview simulation web app powered by real-time AI video agents. Users step into immersive “interview worlds,” each with its own tone, vibe, and difficulty level — from stiff corporate settings to wild startup chaos.

With the help of Tavus AI agents, users experience real-time interview interactions, receive performance feedback based on tone, articulation, pacing, and emotional delivery, and unlock skill trees, XP, and achievements.

It’s not just practice — it’s performance training in disguise.

How we built it

We built MirrorMind as a fully functional web app using:

  • Bolt.new and React.js for the frontend UI
  • WebRTC for real-time video interaction
  • Tavus for generating dynamic AI video interviewers
  • Supabase for backend authentication and storage
  • Emotion & tone analysis APIs for intelligent feedback
  • Tailwind CSS for sleek and adaptive design
  • Python and Node.js to power custom scoring and logic systems

Challenges we ran into

  • Syncing real-time video with adaptive AI interaction
  • Making each “multiverse” world feel unique without overwhelming UX
  • Crafting feedback that felt human and helpful, not robotic
  • Racing the clock after our end-sem exams

What we learned

We built MirrorMind in one intense post-exam week, and that taught us:

  • Time isn’t the problem — alignment is. Our team vision made everything flow.
  • Great design is emotional — people learn better when they enjoy the experience.
  • AI is powerful when combined with empathy, gamification, and storytelling.

Behind the Build

Between exam-induced burnout and back-to-back night sprints, this app became our therapy, our challenge, and our obsession. We divided tasks, aligned features, and debugged like our GPA depended on it.

We didn’t just build a tool — we built a weird little digital multiverse that trains courage.

What's next for MirrorMind

  • Multiplayer mock interviews with real people + AI judges
  • Advanced analytics for body language and facial cues
  • Voice cloning and "Ideal Self" training modes
  • Expansion into other soft skill areas (pitching, negotiation, etc.)

Built With

Share this project:

Updates