Engineering for systems that must work beyond the diagram.

I document decisions, experiments and trade-offs across architecture, systems and AI — with judgment formed in production, not in theory.

Get the next thesis Read the Decision Logs →
  • 14+ years in critical systems
  • Real production
  • ES / EN

Evidence // 01

The proof is in the systems.

  • 14+ years in critical systems
  • 99.9% continuity target on platforms I have operated
  • 100+ incidents diagnosed or managed in production
  • ES · EN technical talks and sessions in both languages

Cases come from enterprise environments and are presented anonymized.

Decision Logs // 02

High-level engineering. In writing.

Not tutorials. Real decisions, real trade-offs, real consequences.

Resilience

Redundancy didn't stop the cascade.

Banking payment system · High availability

Problem
Cascading failures during load peaks degraded the entire platform.
Decision
Stop the propagation: isolate dependencies and introduce circuit breakers.
Result
Critical incidents reduced by 70%.
Read the Decision Log →
Observability Latency diagnosis in distributed systems Architecture Microservices migration
See all Decision Logs

Method // 03

I don't start by writing code.

First I define what must happen, what cannot break, and what the system's contracts are. The implementation can change; the contracts protect the intent.

  1. Vision
  2. Domain
  3. Contracts
  4. Vertical slice
  5. Build
  6. Validation
See the method

Projects // 04

Same method. Different domains.

Right now: a video game in development and two products already running in production. Domains change; engineering principles don't.

Video game · Action roguelite In development

Babel Destiny

An isometric action roguelite about a tower nobody ever finished climbing. You don't win by killing: you win by understanding the tongue of whoever attacks you. A data-driven engine, with every contract written before the first line of code.

Godot 4 · GDScript
Deterministic engine Functional

TrainO System

Problem: generic training plans that ignore real recovery. Decision: a deterministic engine that adjusts every session with explicit rules, not templates. Principle: the system doesn't interpret; it executes.

iOS · Android
Privacy by architecture Functional

VaultOS

Problem: personal knowledge shouldn't have to leave your machine to be useful. Decision: local-first with layered encryption (AES-256-GCM, Argon2id) and on-device AI. Principle: privacy is designed, not promised.

Electron · React · TypeScript
See all projects

Tools // 05New

Does your architecture really hold the SLA it promises?

The availability calculator turns your SLA target into real downtime —per year, month and week— and checks your architecture against the 8-domain checklist. The printable version is free; the PDF and Markdown, with your email.

  • SLA → allowed downtime (year / month / week)
  • How much of your unavailability budget is left
  • 8-domain checklist: PDF, Markdown, or printable
Assess my architecture

Signal // 06

Signal, not noise.

One thesis. One argument. One decision.

Architecture, systems, AI and applied engineering — once every two weeks. No news roundups. No noise.

No spam, never shared with third parties. Unsubscribe anytime.

Profile // 07

Jorel del Portal

Systems engineer specialized in the architecture and operation of critical systems. I build systems, document decisions, and study how engineering principles apply to increasingly complex domains.

"The system doesn't interpret; it executes."