The engineer behind the work.

I'm Shivam Bhadoriya, an AI Engineer at Aaziko Global LLP, based in Navsari, Gujarat. My job title changed in 2026, and so did the work behind it. I write the specification, direct the coding agents that implement it, and review and test what comes back. The judgement calls — what to build, how it should be shaped, whether the result is actually correct — stayed with me. The typing did not.

The clearest way to describe the shift is with one number. In October 2025, AI wrote 0% of the lines I changed. In August 2026, it wrote 99.77% of them. Nothing about that makes the engineering easier; it moves where the engineering happens. A vague ticket used to produce slow code. Now it produces confident, well-formatted, wrong code — much faster. So the specification stopped being paperwork and became the actual deliverable.

What that looks like day to day: I write down the exact behaviour, the field names, the status values and the acceptance criteria before anything is generated. I keep the work in phases small enough to review properly. I read every diff. I write the tests myself, or I make sure the agent's tests fail for the right reasons before I trust them. When a path is going wrong, I stop and rewind rather than argue the build forward. Most of my value now sits in the review, and in knowing which 0.23% of the lines had to be mine.

I did not start here. Before this I shipped full-stack MERN products end-to-end — production Express and MongoDB backends with JWT auth, rate limiting, realtime sockets and structured logging, behind React front ends built to a performance budget. The habits came from there: 70% faster queries through indexing and aggregation, roughly 60% less bandwidth through caching, 98 Lighthouse. Numbers settle arguments that opinions cannot, and that has not changed just because something else is writing the lines.

I studied B.Sc. Information Technology at Vidhyadeep University, and won a pair of inter-college technical competitions along the way. Outside the day job I run AI-PULSE — a YouTube channel that publishes itself, one video and one Short about a trending AI tool every day, entirely on GitHub Actions with no server and no bill. This site is the other proof: a Next.js front end, a MongoDB-backed API, a self-built analytics pipeline and an admin dashboard, all designed and engineered end-to-end and all inspectable from the outside.