About
The Builder
Systems Engineer · Factory Architect · Reasoning Researcher
I design driveline systems for heavy machinery by day. I build autonomous AI factories by night. And somewhere in between, I started asking whether reasoning itself could work like physics instead of language.
Three worlds. One thread: make systems that think for themselves.
"The mind is powerful. But it is peace that powers the masterpiece.
I chose to build a mind that builds for me -
so I could bring vision to my life, not just velocity."
- AdaptiveMind
# Chapter 1 - The Engineer
I work as a Design Engineer in R&D at Ironfield Engineering - one of India's largest agricultural machinery companies. My domain is driveline systems: the mechanical chain that connects engine torque to the wheels that move earth.
I manage Bills of Material across 183+ configurations. I build engineering change workflows inside Teamcenter PLM. I wrote the punch code traceability system - 106 configurations, 71 unique codes, readable by machines and humans. I optimized tire-slip tolerances and pushed acceptable configurations from 70% to 91%.
What it taught me:
Real systems have 183 configurations, not 3. If your architecture breaks at scale, it was never architecture - it was a demo.
# Chapter 2 - The Factory
"What if one prompt could produce production-grade code - no human in the loop?"
That became The Autonomous Factory - a multi-agent AI orchestration system where each LLM plays a defined role in a governed pipeline: one orchestrates, one audits quality, one reviews architecture, one acts as a digital twin of the owner.
Research -> Blueprint -> Engineering -> Deployment -> Forensic Validation
Not a wrapper around an API. A governed system with TDD gates at every phase, cross-project learning, and a Project Learning Ledger that captures every failure so the factory gets smarter with each build.
The factory audited its own code - found issues - and fixed them autonomously. That was the moment it stopped being a tool and started being a system.
What it taught me:
The gap between "AI-generated code" and "production code" is not intelligence. It is governance. Rules. Gates. Verification. The boring stuff is the real stuff.
# Chapter 3 - The Researcher
"Can reasoning exist without words? Without language? As pure dynamics in a field?"
That became io-gita - a physics engine for the mind. Ancient Indian philosophy encoded as forces in a computational landscape. Push it with a question, and it moves. Where it settles is the answer.
The discovery: the shape of the landscape governs the reasoning. Not the words. Not the model. Not the data. The topology. And it works the same way across completely different domains - philosophy, physics, incident response, manufacturing.
It is now live at io-gita.com - a deterministic guidance engine where you describe a dilemma and the engine shows you where your forces naturally lead.
What it taught me:
The most interesting questions are the ones where you do not know the answer yet. And the only honest way forward is controlled experiment, not conviction.
# How I Operate
BLUEPRINT FIRST -> No code without architecture MULTI-LLM REVIEW -> No single model is trusted alone TDD GATES -> If it is not tested, it does not exist CLAIMED != VERIFIED -> Every output gets forensic validation LEARNINGS COMPOUND -> Every failure becomes a rule CAR NOT SPACESHIP -> Build what works. Scale what survives.
# The Thread
Driveline engineering taught me that real systems have hundreds of configurations and zero tolerance for ambiguity.
The factory taught me that AI becomes useful only when you govern it like you govern any engineering process - with rules, gates, and accountability.
The research taught me that the deepest questions about intelligence might not live in language at all, but in the physics of computational landscapes.
Three chapters. One principle: systems that think for themselves need architecture, not magic.