I learn by building.
I’m drawn to ideas that sound interesting but might not quite work yet. Building something is how I find out.
Pranshul Chandhok / 2600th
I’m VP Product & Technology, Interior Company at Square Yards.
I work across AI and real-time 3D, and stay hands-on with the software we build. I’m based in Gurugram.

What I mean by building.
My work moves between product direction, technical architecture, prototypes and implementation. The balance changes with the project. Leading a team does not mean I wrote every part of its software, and a prototype I built is not the same thing as the production platform a team developed from it.
I name those boundaries on the Work pages. My independent projects in the Lab are where I can try an idea, inspect the result and share the code.
2600th is an old handle from my college-era interest in hacker and phreaking culture. It stuck.
Tools in use.
- Browser graphics
- TypeScript, Three.js and WebGPU in Web Ocean 3D. Rapier supports physics in Kinema.
- Applied AI
- OpenAI and Gemini for floorplan interpretation, with ComfyUI for repeatable image and video experiments.
- Product prototypes
- Next.js, Tailwind CSS, AI SDK and fal.ai formed part of the initial PropVR AI MVP. That describes the foundation I built, not the current Craft platform’s stack.
Humanoid Robot Control System
At GreyKernel, I co-invented a system for controlling a humanoid robot using tracked human movement. Indian Patent No. 395331 was filed on 14 February 2019 and granted on 25 April 2022.
My co-inventors are Abhishek Gupta, Devesh Kumar Singh, and Tej Pratap Singh. GreyKernel Technologies India Pvt Ltd holds the patent.
How I work.
- Try the idea.
I often build a small prototype to answer a question. It gives the team something concrete to discuss. - Stay close to implementation.
I test tools and read the code before recommending them. Their limits matter as much as their capabilities. - Pay attention when it breaks.
Slow responses, memory limits, and confusing controls tell me where more work is needed. - Use AI where it helps.
Sometimes a deterministic rule or a database query is the better choice. Some decisions still need a person.
