I design the systems that teams of AI agents use to build, test and ship real software.

You get a working product.
I answer for every line.

See how I workLet's talk

RAFAELRAFAELALVAREZALVAREZ

How a product gets forged

It all starts raw.

An idea, a process that keeps jamming, a half-built repository. Material with potential and no shape yet.

First, the blueprint.

Before an agent touches the code, the system is already written: specs in plain language, atomic issues and acceptance criteria you can measure.

Then, the hammer.

Teams of agents work in parallel on isolated branches. Each one with exactly the context its task needs. None of them improvises.

Nothing ships untempered.

Tests, review and automatic deployment on every change. Autonomy without control is noise; with control, it is speed.

Services

01

Web Development

Full-stack applications built with Next.js, React and TypeScript.

02

AI Transformation

AI inside the business, not next to it: process audits, tailored architecture, integration with your systems and a team that knows how to use it.

03

AI Environments

I bring your development environment into agent-driven work: tools, workflows, context and best practices for the whole team.

04

Products

End-to-end products, from concept to deployment. Desktop apps, dashboards and SaaS delivered in weeks, not quarters.

05

Maintenance

Your system kept current: maintenance of what already works and continuous upgrades to new models and tools.

From idea to production

Diagnosis

A session to understand the business, not just the code. What hurts, what repeats and where an agent truly adds value.

Architecture

I design the whole system: what each agent does, with which tools, what context it gets and who validates its work.

Build

Agents execute in parallel on well-scoped issues. I review, correct the course and keep the bar high.

Delivery

A product in production, documented and tested. Your team knows how to use and maintain it from day one.

Evolution

Models change every month. So does your system: I keep it updated so it keeps performing with the latest.

Selected Work

Every project is proof of how AI is transforming software development.

Skip to the rest of the projects
PHPlay title screen: night-time pixel art city with the logo, a windmill and the farm
Zend Valley: the game world with the house, the field and the BIT-8 drone
Lab with the Zend Lens: PHP code on the left, memory (symbol table and object heap) on the right
PHPlay on a phone: the pixel art title screen
PHPlay on a phone: Zend Valley with the BIT-8 drone

01/ 07

PHPlay

Live

A pixel art RPG to really learn PHP: you program a drone in PHP to bring a farm back to life. 13 chapters, a PHP 8 interpreter written from scratch in JavaScript and a lens that shows memory (zvals, references, objects) step by step.

13
chapters, each with a mentor
0
dependencies: a PHP interpreter written in JS

JavaScript · Canvas · PHP · WebAudio

phplay.dev ↗
ruleaihq.com landing page: “All of your company’s AI. Under control.”
Company map: teams, registered AI tools and the ones used without being registered
Rules panel: which data can be shared with each AI
RuleAI on a phone: “All of your company’s AI. Under control.”
RuleAI on a phone: the company map, team by team, with unregistered AI in sight
RuleAI on a phone: the method, from zero to useful AI step by step

02/ 07

RuleAI

Live

A platform to roll out AI across a company step by step: a map of teams and tools, the AI used without being registered, rules on which data goes to each AI, role-based training and changes with human review. The customer brings the AI, over MCP.

TypeScript · Next.js · PostgreSQL · MCP · Three.js

ruleaihq.com ↗
El Productor, new shoot: goal, a batch of 9 videos and a topic hint, with budget and time shown up front
El Productor, deliveries: 22 videos with their status in Estudio and the revisions each one needed
El Productor in a narrow terminal: a new shoot with a batch of 9 videos
El Productor in a narrow terminal: the 22 deliveries with their status and revisions
In my voice
In my voice
In my voice
In my voice

03/ 07

El Productor

Live

Automatically generated content for RuleAI: a Claude Code harness with its own TUI that writes the script, adds my voice, renders each reel in 3D and schedules it on social media. Batches of up to 9 videos based on my preferences; since it has already captured my taste, I only review.

9
videos per batch, built from my own preferences
1–2
revisions per video, on average
< 1
afternoon of just reviewing for the whole batch

Claude Code · Node.js · Three.js · ElevenLabs · FFmpeg

Instagram @ruleai.hq ↗YouTube @ruleaihq ↗
mcpizer landing page: “Invoking is not acting”, with the commands to validate a policy
An mcpizer decision: allowed, with reason, capability, account, ceiling and file:line:column
The policy line that grants the permission, explained field by field
mcpizer on a phone, top to bottom: commands, what a decision returns and the policy line

04/ 07

mcpizer

In development

An MCP gateway with authorization: for each caller it decides which tools they see, which account runs them and under what limits. Deny by default, never sees credentials, and every decision comes with its reason and the file:line:column that produced it.

TypeScript · MCP · CLI · YAML

mcpizer.vercel.app ↗github.com/rflvz/mcpizer ↗
ChessTogether landing page: “The chess social network” over a 3D board
Chess players near you, with the golden knight on the board
“Every game promotes you”: levels from pawn to king and points per result
ChessTogether on a phone: “The chess social network”
ChessTogether on a phone: players near you, with the golden knight on the board
ChessTogether on a phone: every game promotes you, from pawn to king

05/ 07

ChessTogether

Live

Built in 10 days with a team of 2 devs plus technical direction, planning and organisation.

10
days of development
2
devs + technical direction

TypeScript · Next.js · React · Supabase

chesstogether.app ↗
SeenPDF landing page: “Know when your client opens the proposal. In real time.”
Analytics for a proposal: visits, average time and time per page
SeenPDF on a phone, top to bottom: the landing page, the documents panel and the signals it tracks

06/ 07

SeenPDF

Live

PDF proposal tracking: share a link and see in real time who opened it, which pages they read and how long they spent on each one.

TypeScript · Next.js · React · Supabase

seenpdf.vercel.app ↗
untacit landing page: “Nobody knows how your company works. Now they do.”
The knowledge graph: entities, processes and policies, with a conflict flagged
Conflicts between sources as first-class citizens, with their review queue
untacit on a phone, top to bottom: the landing page, the graph and the problems it solves

07/ 07

untacit

Live

A typed ontological graph of business logic, extracted from code, documents and agentic interviews. Every edge knows where it came from.

TypeScript · MCP · Claude_Code · SQLite · Tauri

untacit.vercel.app ↗

More projects

  1. TeamergyTeam management platform built with TypeScript, Next.js, Python and PostgreSQL.In developmentTypeScript · Next.js · Python · PostgreSQL
  2. StyleClusterVisual style clustering with Python, PyTorch, CLIP and scikit-learn.In developmentPython · PyTorch · CLIP · scikit-learn
TypeScript✕Next.js✕React✕Claude Code✕Cursor✕Rust✕Node.js✕Supabase✕Python✕LangGraph✕PostgreSQL✕PyTorch✕MCP✕Tailwind✕

Writing code is no longer the hard part. The hard part is the judgment to decide what deserves to exist. That is what I bring to the table.

Testimonials

“
Working with Rafa on ChessTogether was one of the best experiences. Solid technical leadership and well-reasoned architecture decisions.
MGMarcos García SánchezDEV — CHESSTOGETHER
“
Working with Rafa changed the way I approach my day-to-day. He set up an AI environment that I ended up weaving into my workflow almost without noticing.
CAClemente Álvarez BoixCOLLABORATOR

Open to Connect

Have a project in mind or want to bring AI agents to your team? Tell me about it.

EMAILrafaceitunoalvarez@gmail.com
GITHUBgithub.com/rflvz
LINKEDINin/rafael-aceituno-álvarez
I reply within 48 hours.

Blog

AI tips, management and training: what I learn building with agents, explained so you can use it tomorrow.

View all posts