Jawahar Naidu — MINDWORLD OS: a personal operating system, Bengaluru to Paris
This site is built as an operating system: it boots on real system data, then travels one continuous camera route through a 3D city of real work. A 90-second recruiter tour, a skill galaxy rendered from the live local Claude Code skill registry (261 skills installed, 13 authored), project rooms with honest metrics, and a hidden terminal. Keyboard: Cmd+K opens the command palette; the backtick key opens the terminal.
THE CITY COMPILES
Every district below is something real — places I studied, teams I worked with, products I shipped. Right now it's a blueprint. Scroll, and the sun comes up on it.
SUNRISE — BENGALURU
First light, 12°58′N. The city that built me: an engineering degree, a first job with real clients, and a community that taught me to run things at scale.
ACADEMIC QUARTER
B.Tech in Computer Science, REVA University, 2021–2025 — honours. First ML models, first deployed apps, first users. The CO₂ prediction project from those years is still public: a model that beat three others on a real dataset, wrapped in an app anyone can run.
CO₂ ML PROJECTCAREER QUARTER
Two working buildings. PocketLite — my first production code: PHP backends and CRM systems. Then PandaECE, where I ran digital strategy for 30+ SME clients and put GenAI into their production workflows: production time down 40%, with a human reviewing everything before it shipped.
CULTURAL QUARTER
Four years volunteering at ISKCON Bengaluru. I led a team of 50+ across events hosting 5,000+ people — zero critical incidents in four years. It's where the Vedic thread in my work starts, and where I learned to stay calm when thousands of things can go wrong at once.
THE CROSSING
2025. One flight, 7,800 km. The city falls away, the day passes below. Everything I'd built fit in two suitcases and a GitHub account.
NEW GROUND — PARIS
Golden light, 49°N. ESSEC Business School, Master in Management — apprenticeship track. Paris added the layer engineering doesn't teach: strategy, financial modeling, and how organisations actually adopt what gets built.
THE FLAGSHIP — ADOPT
The tallest tower, lit like a beacon. Adopt turns any work task into a qualified GenAI use case — fit score, ready-to-paste prompt, adoption guide. Built and shipped as sole product owner, made to mirror a real GenAI-adoption apprenticeship description: the tool that role needs, already working.
OPEN THE LIVE APPHOW IT WORKSTHE WORKSHOPS
Smaller builds, all real. WebForge — an AI agent that generates and deploys websites from a one-line brief, over Telegram. Tara — a voice assistant that listens locally and thinks with Claude. Rta Living — a smart-home platform, in development. And a drafting office where this city itself was drawn, in code.
WEBFORGETARARTA LIVINGTHE INTEGRATION STREET
New city, same method: learn the rules, then build. I work across four languages, and French is the newest — being learned the slow way, in classrooms and cafés. Integration is a build too. It doesn't compile overnight. It's compiling.
THE OPEN PLOT
Dusk. One empty lot, one crane. The next building here is yours — a 12–24 month apprenticeship in France, from September 2026. I ship from week one.
EMAIL MEDOWNLOAD CVGITHUBADOPT — LIVEProject rooms
ADOPT — flagship
The GenAI adoption companion — task in, working practice out.
Sole product owner · designed, built, shipped
Stack: Next.js, React, Claude API, Tailwind, Vercel
SHIPPED 2026 · LIVE · built to mirror a real GenAI-adoption apprenticeship JD
Type a work task. Adopt qualifies it — a fit score for GenAI, a ready-to-paste prompt, and an adoption guide for rolling it into the team's week. The hard part of enterprise AI isn't the model, it's the adoption curve — this is a tool aimed exactly there.
OPEN THE LIVE APPHOW IT WORKSWEBFORGE
A one-line brief in Telegram becomes a deployed website.
Built end-to-end
Stack: Telegram Bot API, Claude (Haiku → Sonnet), Node.js, Vercel
AGENT PIPELINE · brief → generated site → live deploy
Send the bot one line about your business. A cheap model drafts, a stronger model refines, and the result deploys itself. Built to prove an agent can own a full delivery loop — not just a chat window.
WEBFORGE — REPOTARA
A voice assistant that listens locally and thinks with Claude.
Built end-to-end
Stack: Whisper (local), Claude API, Python, macOS
LOCAL-FIRST · audio is transcribed on-device before anything leaves the Mac
Tara sits in the Mac's notch as a quiet orb. Speech is transcribed locally with Whisper; only text goes to the model. Voice UI with the privacy default the enterprise actually wants.
TARA — REPORTA LIVING
A smart-home platform for a real house.
Project — in development
Stack: IoT, Next.js, Home automation
STATUS: IN DEVELOPMENT · honestly marked
A small smart house being wired for real daily use. It's not finished, and it's listed as exactly that — a project in progress, not a claim.
RTA LIVING — REPOCO₂ ML LAB
Vehicle CO₂ prediction — the first deployed model.
B.Tech project, REVA University
Stack: Python, XGBoost, Flask, scikit-learn
4 MODELS COMPARED on a real dataset · public repo · runnable app
An XGBoost model that outperformed three others on a real vehicle-emissions dataset, wrapped in a Flask app anyone can run. Where shipping-over-slides started.
CO₂ PROJECT — REPOTHE DRAFTING OFFICE
The room where this city was drawn — in code.
This site. Designed and built from scratch
Stack: Next.js 16, React Three Fiber, Three.js, zustand, Tailwind
VERIFIED 60FPS at 1440×900 and 390×844 · procedural city, one draw call per district
MINDWORLD OS is itself the résumé line: a continuous camera route, a day-cycle lighting engine shared by every shader, an instanced city, and the skill galaxy running on the live local registry. You're inside the proof.
GITHUB — Jawahars07