Technical Lead · TypeScript & SQL · AI-Augmented Engineering
I turn business problems into working software and clean data by directing AI coding agents and small and large teams (up to 40 people) — and I gate every release.
Technical operations lead with 10+ years in software, 4+ leading teams, from automotive-software implementations to development and leadership. At a B2B SaaS I directed its CRM, WhatsApp automation and 6-VPS production; now I own QA, releases and data migrations, delivering with AI coding agents under verification gates I design. Earlier I led teams of up to 40 at IPN.
Based in Mexico City (UTC−6, no DST) · Works across US Eastern, Central and Pacific hours.
Current role · SaaS for auto repair shops
customer branches whose data (orders, payments, balances, catalogs) I moved from Excel or older systems into the live platform
New customers start working on day one with their full history.
of the records matched the source files in the documented case
Own records · current role · documented case: 8,339 orders from 3 branches, 0 errors, 2026-09
See the migration caseCurrent role · CRM and WhatsApp automation
of manual work removed every month by the CRM and WhatsApp automation I directed (300+ active accounts)
The team spends that time on customers, not on repetitive tasks.
Own records · current role
See it in the CV01What I do
I design the architecture and run it in production: 6 VPS with per-server monitoring, multi-tenant data models and decisions written down with the options I rejected.
Systems that grow without a rewrite and that another engineer can pick up.
I lead development, QA and support teams, from small squads to 40 people: code review, mentoring, playbooks and role-based training.
A team that ships without depending on one person.
I test and approve every ticket before release and check it again in production; I handle incident triage and L2/L3 escalations.
Fewer failures in front of your customers and less rework.
10+ on-site implementations, 100+ remote rollouts and data migrations for 50+ branches; I move data from Excel or legacy systems to production with dry-run, QA and reconciliation.
New customers go live sooner, with no lost data.
CRM, Customer Success and WhatsApp automation with LLM ticket triage; e-invoicing (CFDI 4.0). Two products of my own: VetNova and REYBER.
Less manual work: about 60 hours a month in my current role.
I direct AI coding agents (Claude Code) inside rules, code review and verification gates I design.
More output from the same team without lowering the bar.
02Metrics
Every figure traces to a report, a repository or a console reading.
REYBER · insurance-agency platform (own product)
smaller search index than the alternative design, tested at the scale of a large insurer (10 million policies)
Searches stay fast and servers stay cheap as the data grows.
Benchmark: 492 vs 3,030 MB · 2026-09-22
See the caseCurrent role · QA gate before each release
development tickets I tested and approved before release; 18 sent back for fixes and 30 rechecked in production
Bugs are caught before your customers see them.
QA release gate case · 2026-10
See the caseVetNova · veterinary clinic SaaS (own product)
automated end-to-end tests that hit the real API
Every change is tested on its own: fewer surprises after a deploy.
Counted in the repository · 2026-10-13
See the caseJARVIS · personal voice assistant
to answer common commands by skipping the AI model, against 8.6 s when the model answers
About 3× faster replies and fewer paid model calls.
JARVIS status file · 2026-10-13
See the case03Work cases
Case studies with their decisions, validation and results.
How I gate every release of a B2B SaaS for auto-repair shops twice, in QA and in production, with E2E matrices of 13 to 45 cases and a rule that QA reports and never fixes.
production
Moving the history and catalogs of multi-branch customers out of their old systems with a two-step pipeline, dry-run to QA to production, and a reconciliation against the source at every stage.
delivered
A dry run counted about 3,400 service packages when 11,934 existed. Tracing the mismatch to unordered pagination, and to nulls the first filter dropped, made a bulk cost correction safe.
delivered
04Projects
Multi-tenant SaaS for veterinary clinics, built with AI-directed architecture. Pre-revenue; production-ready; maintenance mode.
production
Multi-tenant core insurance platform for insurers and insurance agencies. A falsifiable prototype on 10 million policies decided the data model before any feature was built.
building
A local voice and text assistant that runs on a laptop CPU with no paid APIs. Safety is enforced by code, not by the model, and a fast path answers common requests without calling the LLM.
paused
How I split engineering work between AI agents by difficulty and cost: a planner that cannot edit, an executor that follows plans exactly, a read-only explorer, and a local panel that shows who is doing what.
production
05How I work
A green check is not proof.
The method and the rules I set for AI agents, in my work and in my own projects.
Read how I work06Contact
Mexico City, working across US time zones. Fastest reply: WhatsApp.
