NJ
Building in public

Naman Jain

I build AI products.
Now I'm building the agents inside them.

Every week, something shipped. Every week, something learned.

I've spent the last 2+ years building next-gen SaaS and AI products — mostly on the frontend, in Angular, React, and Next. I own UI/UX end-to-end, not just components. I've built interfaces for GenAI products. Now I'm learning what runs underneath them — agents, orchestration, and the systems making the decisions.

Scroll

About

I've spent the last 2+ years building next-gen SaaS and AI products — mostly on the frontend, in Angular, React, and Next. I own UI/UX end-to-end: not just components, but the flow, the requirements, the parts that make a product actually work for the person using it.

Along the way I've built interfaces for GenAI products — chat experiences, AI-driven flows, the stuff users actually touch. What I haven't done yet is build what's behind that interface: the agents, the orchestration, the systems making the decisions.

That's what I'm doing now. Learning backend fundamentals, Spring AI, and how agents actually work at the code level — and shipping something new publicly, every week, until I can build agentic products end to end, not just the screens in front of them.

Projects

Built in public. Some are polished, some are this week's experiment — both are here.

Building

Select. Describe. Watch it change.

Click any element on a page, describe the change in plain English, and an LLM rewrites its style and content live — no code, no panel-hunting. Built to explore how natural language can drive direct UI manipulation, the same instinct behind every AI page-builder you've used, stripped down to its core mechanic.

Angular · Groq · postMessage · Iframe DOM bridge

Building

A changelog that writes itself

Most changelogs are an afterthought — bullet points nobody reads. This pulls a week's GitHub commits and has an LLM turn them into something a human would actually want to read. Built mostly to keep myself honest about what I shipped each week.

Why

because “shipped weekly” only means something if there's a record of it.

Node · GitHub API · LLM

Planned

An agent that triages your GitHub issues

Not a chatbot wrapper — an agent that reads open issues, decides what's actually urgent, tags them, and suggests next steps. First real attempt at agent behavior: reading, deciding, acting, not just answering.

Why

this is the first project where the LLM has to decide something, not just generate something.

Agentic AI · Tool use · GitHub API

Let's Talk

Building something interesting, hiring, or just want to talk shop on agents and AI products? I'm around.