DEV Community: Sankalp Kulkarni

Why I Wrote 25 Engineering Rules Before Writing O-AI

Sankalp Kulkarni — Sat, 27 Jun 2026 19:12:34 +0000

Most AI projects start with code.

I started with documentation.

Before implementing O-AI, I spent weeks writing a complete engineering handbook covering architecture, memory systems, multi-agent design, security, testing, deployment, debugging, governance, and long-term maintainability.

Why?

Because AI projects grow fast—and without clear engineering principles they become impossible to maintain.

Writing code is easy.

Designing a system that can still evolve years later is much harder.

The implementation begins now, but the blueprint comes first.

I'm building O-AI in public and documenting the journey along the way.

I'd love to hear how other developers approach large AI projects.

O-AI Development Update

Sankalp Kulkarni — Thu, 25 Jun 2026 19:23:25 +0000

I wanted to share a quick update on how O-AI is progressing.

Over the past few weeks, I've been rebuilding and improving almost every part of O-AI. The goal has always been to create something that's genuinely useful—not just another AI chatbot.

Here are some of the biggest updates so far:

• O-AI Agent Extension – O-AI is evolving beyond simple conversations and can now work more like an AI agent, capable of handling more advanced and multi-step tasks.

• O-Code – A dedicated coding environment built into O-AI, designed to help developers write, debug, understand, and improve code more efficiently.

• Improved reasoning & performance – Faster responses, better reasoning, and a much smoother overall experience than previous versions.

• UI & workflow improvements – A cleaner interface and many under-the-hood optimizations that make O-AI feel significantly more polished.

That said, I don't consider O-AI "finished"—not even close.

There's still a long way to go, and I know there are countless things that can be improved. Every day I find new ideas, fix issues, and think of better ways to build features. I'm treating this as a long-term project, and I want it to keep evolving over time rather than rushing a release.

Building an AI platform like this entirely on my own while being a 16-year-old student has been one of the biggest challenges I've taken on. Balancing school with development isn't always easy, but seeing O-AI improve with every update makes it worth it.

The O-AI Trial Version is getting closer, but before releasing it, I want to make sure it provides the best experience I can build at this stage.

Thank you to everyone who's been following the journey and supporting the project. Your encouragement means a lot, and I'm excited to share more as O-AI continues to grow.

Website: sankalpkulkarni.com
LinkedIn: linkedin.com/in/sankalpkulkarni1012
Email: sankalpkulkarni24@gmail.com

I built a fully local AI assistant at 16 — no cloud, no API keys, runs on your GPU

Sankalp Kulkarni — Mon, 22 Jun 2026 21:44:27 +0000

I'm 16, from Pune, India. For the past couple of years I've been building O-AI — a fully local AI desktop assistant. No cloud. No API keys. No data leaving your machine. Everything runs on your own GPU.

Why I built it

Every AI assistant I tried sent data somewhere. ChatGPT, Copilot, Gemini — all cloud. I wanted something that felt like JARVIS from Iron Man: smart, fast, personal, and private. So I built it from scratch.

What O-AI can do

Core engine:

Runs LLMs fully on-device via llama.cpp / Ollama (zero internet required)
Self-learning core — extracts facts from every conversation and stores them permanently
Fine-tuning pipeline — train the model on your own data, locally

Voice & language:

Voice control in English, Hindi, and Marathi via Whisper (running locally)
Responds in whatever language you speak

Modes:

JARVIS mode — arc-reactor HUD, 4 reactive states, British-male voice, "sir" persona
Take Over PC mode — full desktop automation
Animated floating desktop pet (4 types, draggable, reacts to voice)

30+ automation fast-paths: open apps, search the web, control media, screen vision, run code, edit files, cursor control, social media steps, clipboard ops...

Multi-step agent system: plan → execute → verify loop with 14+ step types (web_search, fetch_url, read_screen, run_code, edit_file, open_social, and more)

Stack

Backend:  Python (Flask IPC + agent core)
Frontend: Electron + vanilla JS
LLM:      llama.cpp / Ollama
Voice:    Whisper (local) + Edge TTS / neural voice
Vision:   PIL + screen capture

The hardest bugs

"Says done but isn't" — Early versions reported success even when an agent step failed. Fixed by building a proper outcome verifier that reads the actual result, not the plan.

The "opens a random video" bug — Asking the agent to play something would open random YouTube videos. Root cause: the plan validator wasn't catching placeholder URLs like [video_url]. Fixed with a universal content guard on all plans.

GPU offloading on Windows — Getting all 32 layers onto the GPU with the right CUDA flags took way too long. Worth it though.

What I learned

Building something real teaches you more than any tutorial. Every bug is a design decision you haven't made yet. If you're not embarrassed by v1, you shipped too late.

Follow along

GitHub: github.com/Shriisoot
Portfolio + TheLab: sankalpkulkarni.com
Instagram: @shriisoot

If you're building something local-first with LLMs, drop a comment — I'd love to compare notes.