Notebook № 001 · Ezequiel Mina lab status: open

ninathelab

laboratory & labrador — both are true

Twenty years of infrastructure and security. Built the platform behind 7M+ users in regulated fintech, from day one to three countries. Now I run experiments here: AI agents, voice, and systems that ship.

See the experiments Get in touch

Exp. 01 Lab log

whoami

The short version, straight from the shell. When it finishes, the prompt is yours.

zeke — bash simulated shell · browser only
# welcome to the lab

Exp. 02 About

A bit about me

Ezequiel Mina

Ezequiel Mina

Argentine engineer · CDMX & Buenos Aires

Recently
VP of Technology at Ualá México through June 2026 — leading Technology, Operations, and Security for the Mexican business, including the on-premise → AWS migration inside a regulated banking perimeter.
Origin
Joined Ualá on day one (Nov 2016) as Manager of Cloud Infrastructure and built the entire platform from scratch — the first large-scale serverless architecture deployed in Argentina, live in 6 months. That foundation ran 7M+ users across Argentina, Colombia, and Mexico on 200+ AWS accounts and 9,500+ Lambda functions.
Path
Started as a pentester at Cybsec (2006), moved through EY, the Argentine Ministry of Foreign Affairs, and 5 years as DevOps Engineer at Globant before betting everything on a fintech startup that was just a PowerPoint.
Now
Going deep on AI and agents — voice agents, automation pipelines, personal AI infrastructure. The domain is the joke and the thesis: Nina the Lab. Laboratory and labrador — that's her up there, on wedding duty.
Languages
Spanish (native) · English (full professional)

Exp. 03 Field notes

Lab log — career

Each role written like an experiment: what we believed, what we ran, what held up under load.

Ualá 9 yrs 8 mos · Nov 2016 — Jun 2026
regulated perimeter — still online
03.1

VP of Technology, México

Mar 2022 — Jun 2026
Ciudad de México
Hypothesis
A regulated Mexican bank platform can move on-prem → AWS and hit group quality without interrupting the business.
Method
Owned Technology, Operations, and Security for the Mexican business — architecture, engineering execution, product growth, and team building in a new market.
Result
Local platform lifted to the same infrastructure bar as the rest of the group while staying inside a regulated banking perimeter.
Cloud Migration Regulated Environment Engineering Leadership Product Strategy Team Scaling Fintech
200+ accounts · 3 countries
03.2

Head of Technology

Jul 2021 — Mar 2022
Argentina
Hypothesis
Multi-country fintech infra can scale inside Banxico / BCRA / SuperFinanciera constraints without heroics.
Method
Owned the full infrastructure estate as Ualá expanded across Argentina, Colombia, and Mexico.
Result
200+ AWS accounts, 9,500+ Lambdas, 7M+ users across 3 countries — all inside a financial-services regulated perimeter.
200+ AWS Accounts 9,500+ Lambdas 7M+ Users · 3 Countries Regulated Fintech Kubernetes Terraform
day one · PowerPoint → prod
03.3

Manager of Cloud Infrastructure

Nov 2016 — Aug 2021
Argentina
Hypothesis
A serverless-first stack can carry a regulated fintech from zero to production in months, not years.
Method
Day one hire. Designed and shipped the entire platform on AWS Lambda + API Gateway + DynamoDB before serverless-first was the regional default.
Result
Production in 6 months. First large-scale serverless architecture deployed in Argentina — the foundation that later ran the multi-country estate.
0-to-1 Architecture First Large-Scale Serverless in AR AWS GCP Terraform Jenkins
Earlier experiments 2006 — 2016
03.4

DevOps Engineer

Globant

Oct 2011 — Oct 2016
Buenos Aires · 5 yrs
Hypothesis
Enterprise infra gets better when every decision carries an infosec lens.
Method
Designed and scaled infrastructure for clients like Scripps Networks, Southwest Airlines, and WPP.
Result
The exact combo a fintech startup needed in 2016 — build fast, break less.
DevOps Infrastructure Cloud Monitoring Linux
03.5

IT Security Coordinator

Cancillería Argentina · Ministerio de Relaciones Exteriores

Jan 2009 — Dec 2009
Buenos Aires
Hypothesis
Owning the program beats only executing tickets.
Method
Pentesting, server hardening, IDS and firewall deployment, ongoing audits.
Result
First security program owned end-to-end — the shift from operator to owner.
Penetration Testing Firewalls IDS IT Audits
03.6

Security Consultant

Ernst & Young

Jul 2008 — Dec 2008
Buenos Aires · 6 mos
Hypothesis
Standards and checklists only matter if they survive real audits.
Method
SAP security research for AySA, IT auditing for Warner Music, security standards development.
Result
Consulting reps across enterprise security — useful later for regulated fintech.
SAP Security IT Audit Security Standards
origin story · how things break
03.7

Security Consultant

Cybsec S.A.

Sep 2006 — Jul 2008
Buenos Aires · 1 yr 11 mos
Hypothesis
You build safer systems after you’ve broken enough of them.
Method
Network, web app, and wireless penetration testing.
Result
Where it started — the muscle memory for how things actually fail.
Penetration Testing Network Security Web App Security Hardening

Exp. 04 Specimen shelf

Things I've built on the side

Side labs that turned into real systems. Filed by family, tagged by how far each one made it out of the petri dish.

production prototype experiment
★ Featured specimen · P-02 · production

Karla — AI Voice Agent

Voice AI for Mexican logistics · Claude · Deepgram · ElevenLabs · RingCentral

<1.5s
STT → LLM → TTS round trip
200K
calls / month designed
~$0.020
per call

Production voice agent that calls truck drivers in colloquial Mexican Spanish to validate cargo trip data — FSM in Redis, tool-arg pinning against prompt injection, and a latency budget tight enough to feel human.

Open specimen → PoC on GitHub ↗

Applied AI — in real operations

P-01 prototype

Autek AI

Agent-driven regulatory audits for Mexican fintech — with citations regulators can actually follow.

LangGraphRAGQdrant
Open specimen →
P-06 production

Maschitto

Rental ops platform where the transaction classifier learns patterns instead of re-calling the LLM every time.

ReactSupabaseGroq
Open specimen → Live site ↗

Personal AI infrastructure

P-03 production

Clawdia & openclaw

A serious personal AI stack — agents, memory, tools, and a phone number — on a single VPS.

pgvectorMCPSIP
Open specimen →
P-04 prototype

Finance Advisor

Portfolio agent under citation discipline: cite a source, or a deterministic formatter takes over.

Schwab APIopenclawPython
Open specimen → GitHub ↗

Chaos lab

P-05 experiment

Timba 2027

An autonomous political persona living on X — messy on purpose, instrumented on purpose.

LLM agentpersonaX API
timba2027.com ↗ GitHub ↗
P-07 experiment

justo-mx-api

Give an AI agent the ability to grocery-shop at Jüsto (MX) — CLI, JSON out, human confirm before checkout.

Python CLIagentsopenclaw
GitHub ↗

Exp. 05 Skills

What I reach for

AI & Agents

LLMs · AI Agents · LangGraph · RAG · Vector DBs · Claude · Gemini · Tool Use · Voice AI · ElevenLabs · Whisper

Cloud & Infrastructure

AWS · GCP · Kubernetes · Terraform · CloudFormation · Serverless · Docker · Linux

CI/CD & Databases

Jenkins · AWS Code Tools · PostgreSQL · MySQL · DynamoDB · RDS

Security

Information Security · Penetration Testing · Firewalls · IDS · Cloud Security · DevSecOps · Hardening

Leadership

Engineering Leadership · Team Building · Platform Reliability · Incident Management · Fintech · Startup Scaling

Exp. 06 Lab desk

Leave a note in the lab.

File a ticket for roles where I can ship AI with customers and own outcomes — plus advisory work in fintech, infra, and LatAm. Nina reviews all inbound treats.

Open a ticket GitHub LinkedIn X