AI / ML Engineer
Pol Fernández
Building intelligent systems
for a smarter tomorrow.
- Building software
- +5
- Systems in production
- 20
- Working languages
- DE EN ES
About me
I’m Pol.
AI and software engineer in Zurich. I build systems that take a decision, call a tool and leave a record of what they did. Most of the work is the unglamorous half: the integrations, the permissions, the retries, the evidence afterwards.
I grew up in Barcelona, I work in German, English and Spanish, and I learn on my own hardware before I touch anybody else’s production.
- Based in
- Zurich, Switzerland
- Focus
- AI orchestration and automation
- Stack
- Python, TypeScript, Postgres, Linux
- Languages
- German, English, Spanish
What I work with
Languages
- Python
- TypeScript
- SQL
- Bash
AI
- LLM orchestration
- Multi-agent systems
- Retrieval
- Evaluation
Platform
- Docker
- Kubernetes
- Postgres
- Linux and systemd
Delivery
- Azure
- GitHub Actions
- Playwright
- Proxmox
Featured project
Pharma and lab robotics
Multi-agent orchestrator for test automation
A test suite written years ago, in a tool that is being retired, covering machines that handle patient samples. Seven agents share fifty two tools and three model backends. One writes orders, one reads logs, one reviews pull requests, one runs the suite.
The agents are not the interesting part. The interesting part is that every run reports what it cost and what it changed, which is the only honest way to tell which agent earns its place and which one is a demo.
- Specialised agents
- 7
- Tools
- 52+
- Model backends
- 3
- UC-01Order generator
- UC-02Log management
- UC-03Deck visualizer
- UC-04Protocol parser
- UC-05Pull request reviewer
- UC-06Knowledge consolidation
- UC-07Test orchestration
Other work
No client names. The sector says enough.
-
Defence and optronics
Retrieval over documents that cannot leave the building
Air gapped, every answer pointing at the page it came from.
-
Health and wellness
A CRM with the access rules inside the database
Row level security in Postgres, with a test per role that proves it.
-
Pharma and lab robotics
Quality gates that can stop a release
Containerised runners and a virtual machine snapshot per run.
-
Research and prototyping
3D scanning and a robotic arm
Geometry in, motion out, and a simulator so nothing meets a wall twice.
-
My own systems
Agents that report what they cost
One reads the morning signals, one moves my hours between two employer systems.
The lab
I don’t learn this on somebody else’s production.
There is a small rack in my flat in Au running about twenty services: single sign-on in front of all of them, a proxy issuing its own certificates, monitoring that reaches me on Telegram, backups in six layers. It exists because the fastest way to understand how something breaks is to be the one who fixes it at eleven at night.
- Services
- Around 20, each in its own container
- Access
- Single sign-on in front, nothing exposed by accident
- When it breaks
- It says so, on Telegram, before I notice
- Backups
- Six layers, tested by restoring them
Let’s work together
Have a problem worth solving?
Tell me what is not working. I answer within one working day, and if it is not something I can help with I say so in the first reply.
Based in Au on Lake Zurich. Working in German, English and Spanish.