19, Building Fast

Theory is safe. Production is the truth.

Former electronics engineering student, 19. Shipping AI products: Voice AI for Dubai real estate, dev tools for coding agents. Previously: AgTech hardware startup that reached $500k+ pipeline in under 9 months.

View Work Get in Touch

Currently shipping Grov. Join waitlist: stefvirgil2006@gmail.com

Voice AI Dev Tools EF Zurich Full-Stack

Current Focus

Two Projects, One Pattern: Build Fast, Ship Faster

Casius — Voice AI for Dubai Real Estate

Live

The problem: Dubai agents waste 6+ hours daily answering repetitive questions from unqualified buyers. They miss calls during showings. Serious buyers with $2M+ budgets get voicemail.

The solution: AI that answers in <60 seconds, speaks English/Arabic/Russian/Hindi, knows the agent's entire inventory + Dubai geography, scores leads in real-time.

Response Time

<60s

Languages

Time to MVP

Weeks

Technical: Custom voice pipeline (not a wrapper). Meta campaign integrations, property inventory sync, lead qualification logic, WhatsApp routing. Next.js frontend, Node/Python backend, MySQL, self-hosted VPS.

Status: Product finished. First agency in pipeline.

Visit Live

Grov — Collective Memory for AI Coding Agents

Launching

Hooks into Claude Code sessions, captures reasoning, injects relevant context into future sessions. Your AI stops re-exploring the codebase every time.

Exploration Time

1-2 min

Token Burn

<2%

Time to v0.1

48 hrs

Before: 10+ min exploration, 7%+ tokens burned, 3+ explore agents per session

How it works: SQLite storage → GPT-3.5 extraction → semantic filtering by project → contextual injection on next session. Zero friction, works invisibly in background.

Status: v0.1 ready. Launching on npm/GitHub/HN/Product Hunt soon.

Visit Site

"Build fast, ship faster, learn from failures, repeat."

Portfolio

What I've Built

SSA — AgTech Hardware + ML

Wound Down 2025

Solar-powered IoT devices running custom neural networks for autonomous irrigation/fertilization on premium crops (blueberries, etc.).

Pipeline Value

$500k+

Hardware Margins

54%

Timeline

<9 months

What happened: Secured pilot farms across Romania, Germany, Austria with zero deployed product in less than 9 months. Negotiated with manufacturers. Had connections at Driscoll's ($4B+ revenue/year). Ran out of money before shipping.

What I learned: Optimize for tight feedback loops. Revenue is the only unit test that passes. I bootstrap to maintain control of the stack, and I target markets where I can ship, break, and fix in the same day.

Technical: Swift mobile app (full-stack), PyTorch neural networks, custom hardware design. Publication + technical details coming.

EF Builder's Retreat

Nov 2025

Entrepreneur First's intensive builder program in Zurich. 1 of 23 selected (100+ applicants), youngest participant.

Duration

3 Days

Selection

1 of 23

Award

Peer Win

Project: Browser extension allowing SMB owners to modify website frontend via plain English → push directly to production.

Timeline: Flew solo Friday, hacked Saturday, presented Sunday morning. Won peer selection as 2-person team. Also got cool stickers and merch.

Base Jump Arena

Nov 2025

Weekend-long hackathon. 1 of 7 Romanians selected, 1 of 30 total participants.

Participants

Duration

Weekend

Project: Dimini — AI-powered therapy assistant that creates live semantic relationship maps during sessions. Uses GPT-4 for entity extraction and OpenAI embeddings for semantic similarity. Real-time visualization with React Force Graph.

Stack: Next.js, FastAPI, Supabase (PostgreSQL + Realtime), pgvector, GPT-4, OpenAI Embeddings.

Recent Activity

Build Log

Casius

Voice AI for Dubai real estate. Custom voice pipeline, 4 languages, lead qualification. First agency in pipeline.

Grov

Idea to working product in 48 hours. v0.1 ready, launching on npm/GitHub/HN/Product Hunt soon.

Nov 2025

EF Builder's Retreat: 3-day sprint from idea to working prototype. Won peer selection as 2-person team.

Nov 2025

Base Jump Arena: Built Dimini, an AI therapy assistant with real-time semantic mapping. 1/7 Romanian, 1/30 total selected.

Thoughts

Writing

Act Today

I spend an unhealthy amount of time thinking of the future, envisioning the life I desire. But I've realized: the present is far more important than what was or will be. No human can alter time, so learn from the past, and let your decisions from TODAY shape the future.

Your relationships, health, and career will be far more affected by decisions you take today than by anything in the past. Don't act in regard to the past or future—act in regard to today, in calibration to the future you're designing and patterns you've learned. Otherwise, you're just "being busy," not "building."

Divine Burnout

To burn out trying to become something greater is divine. To fade safely as someone you never were is a tragedy.

The Founder's Loop

The emotional waveform every real founder rides:

I'm going to change the world → I can't even get out of bed → Everyone else is ahead of me → I'm a fraud → Wait... what if this works? → Back to #1

That loop never ends. You just get better at surfing it. Feeling lost and questioning yourself isn't a bug—it's the cost of playing the founder game. Anyone who doesn't feel insane a few times a week is playing too small.

Hold Them Accountable

You will lose old friends, not because you're too busy—because they no longer understand the game you're playing.

Hold your friends accountable. Don't let their potential die inside them. Remember, the wheel is round.

Obsession and Sacrifice

Founding a startup is the combination of obsession and sacrifice. You work 7 days a week, fueled only by vision and faith.

I recently read that a bored mind is like an opened door for the devil, and I couldn't agree more. What I've come to realize is that I'm happier the more I work. Humans need challenge. We were built to think, build, solve.

If you have a dream but you're not willing to work on it or fix your systems, you're just disappointing yourself daily. A focused life is not a better life, but the only life one with big aspirations should live.

If you've worked 10-hour days and felt an overwhelming sense of accomplishment at the end, you're doing it right. Keep building.

Stay Foolish

Experience is often just a collection of reasons why something won't work.

The world belongs to those who don't know their limits. Once you learn "how things are done," you stop inventing how they could be done. Stay foolish. It pays better.

Pattern

Why Any of This Matters

I ship.

Casius went from idea to finished product in weeks. Grov went from "this would be useful" to working v0.1 in 2 days. SSA reached enterprise pilots and $500k+ pipeline in under 9 months. Not ideas. Not plans. Shipped products with real traction.

I do the whole stack.

Trained neural nets from scratch. Built full-stack apps. Negotiated supply chains. Closed enterprise deals. Shipped hardware. Shipped software. Shipped AI products. I don't hand off—I execute across the entire spectrum.

Technical range: Full-stack (React, Next.js, Node, Python), ML (PyTorch, custom model training), hardware-software integration, voice AI pipelines, dev tooling, infrastructure (VPS, Docker, deployment).

I fail forward.

SSA didn't work out. We ran out of money. But I learned how real companies operate, how real fundraising works, how real enterprise sales close. Every failure is a data point. Now I'm building faster and leaner.

Execution speed.

MVP-grade products in 24-48 hours. Production-ready systems in weeks. I move faster than most people think, which matters when you're building in competitive markets.

Technical Foundation.

Raised on C and C++, now focused entirely on shipping with Node, Python, and React. I traded manual memory management for velocity, but the systems-level perspective remains.

Education: M1 Baccalauréat with 8.50/10 average across Math, Programming (C++), and Romanian — achieved by studying only in the last 2 months.

Oh, and the logo in the header looks pretty interesting if you hover on it...

Get in Touch

Let's Talk

Open to partnerships, pilot customers, technical collaboration. If you're building something ambitious or need help shipping fast, I want to hear about it.

Send Email Download One-Pager