I am Jay Kadam, an AI/ML engineer in Valsad, India. Most of my work is Python: LLM agents, computer vision, the Flask backends and infrastructure that keep them running once they are no longer a demo.
The job title varies by who is asking: AI engineer, machine learning engineer, AI developer, LLM engineer. I use AI/ML engineer, and the work underneath is the same: models and agents that have to hold up in production, not in a notebook.
I am a core engineer on TestGrid's CoTester, an autonomous testing agent that turns plain-English intent into executable tests and repairs its own output before anyone sees it. It has shipped across nine versions to real customers, which is the part I find worth talking about: an agent behaves very differently once people who did not build it start relying on it.
Alongside that I have trained vision models to find interface elements from pixels when the DOM offers nothing stable to anchor to, and built the matching logic that keeps a test suite running when an interface moves underneath it.
My MBA is in Data Science, from Amity University, Noida, in 2024; the thesis forecast employee churn in South Gujarat's IT sector and set out retention strategies against it. The full record of roles, modules shipped and study is on the experience page.
Most writing about agents describes what they can do in principle. Less of it describes what breaks, what it costs per conversation, or which failure modes only appear in month four. That gap is what I write into.
I compete in bodybuilding, in Men's Classic Physique and Men's Physique, and I run marathons. Both are the same discipline as shipping software that holds up: unglamorous repetition, measured, over a long enough period that the results are not luck.