Most product problems worth solving need data. Most data problems worth solving need product judgment. I work at that boundary.
AI with judgment
I'm a payments product manager who decides where a model earns its place — and where it doesn't. Deterministic where money moves; models at the edges; evals as the only definition of done. Hallucinations aren't wished away: they're identified, gated and mitigated.
- A payment router that keeps the model out of the decision — an enum-constrained decline normalizer with a safe fallback is the only place it runs.
- A golden-set eval with a hallucination gate that turned “the prompt feels better” into a scored regression.
- A remote MCP server where row-level security is the only authorization an agent gets.
- Guardrails that corrected me — a broken API rewrite, a mislabeled issuer, two wrong assumptions of mine, all caught before shipping.
Trajectory
Fifteen years across the full technology stack — support, enterprise networks, Linux infrastructure, QA, full-stack development, technical leadership, platform product ownership since 2023 and global product management since 2024. My domain is payments — checkout, recurring billing, PSP and gateway integrations over REST APIs and webhooks (Stripe, Adyen, PayPal), event reconstruction and reconciliation — and my method is AI with judgment: I decide where a model earns its place and where it doesn't, keep the decisions that move money deterministic and auditable, and treat hallucinations as something to identify, gate and mitigate — never assume away — with evals as the definition of done. Today I run Wolters Kluwer Passport for Marsh, operating daily in English with business stakeholders in Houston, New York, Toronto and London and vendor teams in Houston, Toronto and Pune. On my own stack: a payment routing orchestrator (deterministic PSP routing on empirical evidence, LLM only at the edges, evals, MCP server) and nutri., a production app with RAG, semantic search and a self-built remote MCP server, every LLM change gated by evals. MSc in Data & Information Science in progress (INFOTEC).
- 2010
Support
Banking application support under ITIL v3 — Banorte.
- 2014
Enterprise networks
DNS, DHCP, DMZ, IPv4 — Iusacell Enlace / TotalPlay Empresarial.
- 2015
Quality assurance
Process validation for field operations — Dell Services / NTT Data.
- 2016
Linux infrastructure
Zimbra, Odoo, Zammad, Asterisk, Ubuntu Server — ARIHUS.
- 2019
Leadership
Led a 14-engineer support team — NTT Data.
- 2021
Platform operations
OTT incident & SLA management: HBO Max, Paramount+ — IZZI Telecom.
- 2022
Full-stack & integrations
Salesforce platform owner — payment and money-movement flows end to end (PayPal, Stripe, Adyen on AWS) — Protégitas · LegalBrave.
- 2024
Product management
Wolters Kluwer Passport, global — Marsh.
- 2026
Applied AI in production
RAG, semantic search, remote MCP server, eval-gated LLM changes — nutri., a personal product built and operated solo.
In numbers
- 15+
- years across the technology stack
- 5
- cities: Houston · New York · Toronto · London · Pune
- 2
- daily operating languages
- MSc
- Data & Information Science, in progress · INFOTEC
Contact