Product & Growth | Market Research to 0-to-1 Growth to Shipped AI Products
A reusable knowledge-base engine forked across five domains, from peptide therapeutics to financial compliance. A 14-tool Python substrate validates every build, and an adversarial pass signs it off before anyone reads it.
Influencer audience intelligence across Instagram, X, YouTube and TikTok. Authenticity scored across five dimensions, delivered as a white-label client report while the lead is still warm.
B2B targeting intelligence. Discovers, scores and exports ranked prospect lists across five weighted dimensions. No LinkedIn account, no scraping, no account risk.
I do not just advise on systems, I run them. These are the autonomous agents and evaluation tools I built and operate across my products and engagements. They are how one person ships work that used to take a team.
A coordinated set of agents, each given one narrow mandate, that run country and competitor studies in parallel and return sourced, confidence-rated briefs. This is the system behind the three-country market-entry programme.
Every knowledge build passes a maintenance gate that runs index, lint and evaluation together, then an adversarial pass that tries to prove the content wrong before a reader sees it. It has caught fabricated facts that passed every earlier check.
Load-bearing facts in each domain are pinned to written assertions, 9 to 20 per domain, so a silent corruption flips a check red instead of reaching a reader. The same discipline runs my LLM evaluation and model-selection work.
Output quality across my products is measured by blind-judge evaluations against a written rubric, alongside latency and cost per unit, not by preference. Model selection is a score, not a hunch.
O.P. Jindal Institute of Technology, 2012–2016