Jay Kadam · AI/ML EngineerFrom a rough idea to something people rely on.

01What I do

Short version: give me something boring, repetitive or half-impossible, and I will build the thing that does it.

• Agents that ship

  Not demos. CoTester runs in production across nine versions for real customers. I know what breaks when an LLM meets the real world, and how to keep it from breaking.

• Plain English in, working output out

  Intent goes in, validated and executable steps come out, and the agent checks and repairs its own work before anyone sees it. That is the pipeline I built at TestGrid.

• AI inside what you already have

  Models and agents wrapped into secure APIs and services a team can call on day one. Your app, your platform, your CI pipeline, your internal tools. No rewrite.

• Python for anything repetitive

  Scraping, reports, document and file wrangling, data pipelines, glue between systems that were never meant to talk. If a person does it by hand every week, it can run on its own.

• Vision and ML, end to end

  Computer vision, forecasting and prediction, owned from labelling the data myself to running the model in production.

• One person for the whole thing

  Idea, prototype, production, deployment, monitoring, cost tuning. No hand-offs, and no gap between the person who designed it and the person who runs it.

02Selected work

All work

03Experience

1. 2021 to now

   AI / Machine Learning Engineer

   TestGrid Labs Inc.

   Core engineer on CoTester. Shipped the multi-tenant agent platform, the natural-language-to-execution pipeline, self-validation and auto-correction, and the telemetry that tunes cost per conversation.

2. 2022 to 2024

   MBA, Data Science specialisation

   Amity University, Noida

3. 2021

   Data Science Intern

   BrainyBeam Technologies

   Large-scale data cleaning with Apache Spark and PySpark, and logistic regression implemented from first principles.

4. 2018 to 2022

   B.E. Computer Engineering

   Gujarat Technological University

Full experience

04Writing

• Sept 2026

  What breaks when an LLM agent meets real users

  A field guide to the failure modes that only appear once people who did not build the agent start relying on it, one per section, with what worked against each.

• Sept 2026

  CoTester: building an LLM agent that checks its own work

  Why generated steps break, how self-validation catches it, and what nine versions in front of real customers actually taught.

• Sept 2026

  Cutting cost per conversation in a production LLM agent

  Telemetry worth collecting, where model selection actually pays, and what to measure before optimising anything.

All writing

05Toolkit

Full toolkit

06Elsewhere

• Emailjay@kadamlabs.com
• GitHubjay6430
• LinkedInkadamjay

Get in touch