FullStack · AI and LLMs · Agent-driven automation

I am a FullStack Developer, and I also work with AI, LLMs and agent-driven automation.

Right now I work on React and Next interfaces · APIs in Python and Node · Linux, Docker and TLS · AI, LLMs and agents

  • Location: Málaga, Spain
  • Availability: Immediate
  • Work mode: Remote or hybrid (Málaga)
  • Work permit: Right to work in the EU
  • Currently: Blockchain Developer at BitRise21
  • Languages: Spanish (native) · English (professional, B2)

I build the whole product: the interface in React and Next with TypeScript, and the APIs behind it in Python with FastAPI or in Node. I deploy it myself on Linux and Docker, on infrastructure I own, and I bring in AI and LLMs where they actually solve something.

Single-column resume, parseable by applicant tracking systems.

whoami

Adrián Nieto — FullStack Developer

stack --top

react · python · docker · llms

status

self-hosted · all systems green

Click and type. `help` lists the commands · Tab completes · Esc leaves.

Full-stack, in four areas

  • Frontend

    • React
    • Next.js
    • TypeScript
    • Tailwind
  • Backend

    • Python
    • FastAPI
    • Node
    • PostgreSQL
  • Infra / DevOps

    • Linux
    • Docker
    • Caddy · TLS
    • WireGuard
  • AI & LLMs

    • LLMs
    • Agents
    • PyTorch
    • Neural nets

Published work

Projects

What each project does and how it is built: the stack, the decisions, and what is actually running behind it.

  • Screenshot of the interface of Botanium

    Botanium

    Live demo

    Garden, game, and a whole botanical atlas in your pocket

    It tells you when each job is due on your plants, lets you collect species like a game, and is heading towards spotting pests with models of my own.

    420,528species in the atlas, searchable offline

    • Node 24
    • node:http
    • node:sqlite
  • Screenshot of the interface of AgenticPromptMethod

    AgenticPromptMethod

    Archived

    A methodology for automating code development under supervision

    The work is split into phases, and between one phase and the next automatic gates validate what was done before anything moves on.

    Supervisedagent-driven development: phased, with automatic gates

    • Claude Code (skills)
    • Node.js >= 20
    • node:test
  • The long gamea poker AI trained on what fits in a house

    DeepPlay

    In development

    An AI that learns to play poker, on modest hardware

    A long-term project. The reference is Pluribus, trained on a modest server: proof that this is reachable without a datacenter behind it.

    • Python
    • PyTorch
    • OpenSpiel
  • Screenshot of the interface of ArmyTrading

    ArmyTrading

    In development

    An algorithmic trading platform that always trades in simulation

    It ingests the market in real time, stores millions of candles, tests strategies against the full history and executes signals under risk limits. The real-money mode ships switched off.

    ≈18,500×faster query: 44.7 s → 2.4 ms

    • Python 3.12
    • FastAPI
    • asyncio + uvloop

Infrastructure I run

This runs on infrastructure I own

No managed hosting, no platform with a deploy button. The site is served from a machine I administer myself, and the demos from a home server on the far side of an encrypted tunnel.

System status

LIVE

Operational

  • public web · TLS~12 ms
  • encrypted tunnel~20 ms
  • containersunprivileged
  • monitoringevery 60 s
  • Published demos

  • Botaniumno reading
  • ruskyno reading
Containers in production
25

Every demo isolated, unprivileged, with egress locked down.

Up for
weeks
System load
Server
16 threads · 46 GB · 1 TB NVMe

The heavy lifting for the demos happens here, not on the VPS.

Releases
TLS + atomic symlink

Automatic certificates and a timed rollback path.

Tunnel
WireGuard · ~20 ms

Direct link, without a single port opened on the router.

The line in the panel header is an illustration, not a measurement.

The whole chain, link by link

In progress

Right now

  • DeepPlay, the long one

    An AI that learns to play poker on its own. It is the project I keep coming back to and it is a long game: it trains on the hardware I have at home, no datacenter and no rented GPU, and it moves at the pace that allows.

  • Live demos, on my own server

    Two applications are already running on my own machine, each in its own container and behind an encrypted tunnel, so they can be poked at instead of read about. The rest are landing one by one.

  • This site, and what holds it up

    I run the portfolio and its infrastructure as a product: atomic releases with a rollback path, certificates that renew themselves, monitoring, and a verification suite that runs before anything ships.

How all of this is wired up

Let’s talk

Contact

The address is plain text and works without JavaScript. The subject line is pre-filled; I reply in English or Spanish.

admin@itstatinxd.dev