AI Native Product Lead / Consumer Identity & Customer Core Data / Capital One

I buildAI productspeopleactually use

Platforms, agents, evaluations, memory, and the governance underneath. I write the first commit myself and stay until it is funded. This site runs on one of my ideas. Check your warrant.

GitHub Substack LinkedIn

Shah Baig / spec

RoleAI Native Product Lead

OrgCapital One

FocusPlatforms, agents, evals, trust

Patents40+ filed

BaseNew York & California

RangeWherever the work is

NYEastern time

CAPacific time

StatusShipping

2,000 platform users 60% faster to market 70% lower model spend 57% fewer hallucinations 40+ patents filed 1,500 people trained 78.2 workshop NPS pip install rift-eval

Fig. 001

Selected proof

          012,000users
I founded Capital One's internal agent orchestration platform and wrote the first commit myself. A guerrilla project became a funded product: 50+ contributors, a team of 8 engineers, and a 60% cut in time-to-market.
Read the story

          0235%cost drift
"You upgraded your model. What broke?" Rift is my open-source CLI that answers with paired significance tests and cost per correct answer on every verdict. It caught one silent release raising cost per correct 35%. Ships on PyPI.
See the projects

          0340+patents filed
Filed patents in agentic AI and identity, including an experience layer that tells human users from automated agents and governs what each one can access.
How I think about trust

        

Fig. 002

The trust layer, as filed

One idea from a 40+ patent family, redrawn. The actual filings have more arrows. Most are not public yet. One is: US 2025/0119430 A1 How it plays out on you

Fig. 003

Off the record

Between New York and California

The numbers above are the work. Here is the part that does not fit a spec sheet.

I build things nobody asked for, and I finish them. Before agents I did product for connected appliances at Whirlpool and enterprise platforms at the University of Rochester. I reverse-engineered my own voice recorder because I did not like where the audio went. I finished a master's at Georgetown this May, 3.93, while shipping everything on this site.

The question I keep coming back to is what an agent should be allowed to do on your behalf. It is not the only thing I build. If you are shipping serious AI, I want to meet you. I split my time between New York and California, and I will go where the work is.

Shah

Fig. 004 / Now

I build AI products end to end

Platform, agents, evaluations, memory, governance. I take AI products from first commit to funded, and I go where the hardest problems are. Location is negotiable. The problem is not.