DANGALCAN / PORTFOLIO '26
--:--:-- · CET

PhD student · software engineer

Daniel
Galván
Cancio.

Black and white illustration of a hooded coder in sunglasses typing at a terminal
@DANGALCAN PhD student
Based
Spain · CET
Focus
SLAs · LLM infra
Stack
TS · JS · Python
Status
Open to collaborations
On GitHub
Since 2021
Lang
ES · EN · FR

// 01 · Intro

I build software that keeps its promises, from service level agreements to the infrastructure that serves language models. My final degree project, SLA monitoring for chains of services, won the ETSII 2025 award for the best final degree project in Software Engineering. Right now I work on LEIA, AI assistants for software engineering training, and on EOSC-ARENA, a Horizon Europe project on agentic AI for research.

Software engineer SLA governance service chains LLM infrastructure AI agents Kubernetes open source

Peer-reviewed papers · JCIS 2025 · JISBD 2026
2
CERN openlab · Lunar Lander RL challenge
1st
Best SE final degree project · ETSII 2025
1st

// 02 · Selected work

Repositories
& experiments.

Selected public work
github.com/Dangalcan

The LLM-as-a-service platform of the University of Seville. I am its lead developer and designed its architecture, and I also maintain it and run its deployments.

  • JavaScript
  • Updated · Sep '26

Git-based declarative configuration for the catalogue of the CERN Tape Archive, built during my CERN openlab summer studentship. A reconciler detects drift between the desired state kept in Git and the live catalogue, and applies the difference.

  • Python
  • CERN openlab 2026
  • Zenodo report

My entry to the reinforcement learning challenge of the CERN openlab Summer Student Programme. It finished first on the leaderboard with a mean return of 289.28 over 100 fixed seeds, trained with PPO and refined with an evolution strategy.

  • Python
  • 1st place
  • Updated · Jul '26

Cooling control for the LHCb data centre at CERN, built for a CERN challenge where it finished second.

  • Python
  • Updated · Aug '26

A Node.js module that configures API gateways from SLAs written in SLA4OAS. I implemented its plugin system and developed its plugins.

  • JavaScript
  • Updated · Feb '26

An extension of Bluejay for SLA metrics on chains of services. I added the lucidchain event type, which audits six issue indicators on Redmine and iTop.

  • JavaScript
  • Updated · May '25

A fork of the LEIA runner that adds a new model provider backed by the ALMA LLM cluster.

  • JavaScript
  • Updated · Oct '26

A repository of feature models in UVL format, integrated with Zenodo and flamapy and developed by DiversoLab. In my fork I extended its CLI so the tests run on Firefox and Chromium and added an integration with Selenium Hub. That work earned me honors in the course.

  • Python
  • Graduated with honors
  • Updated · Oct '25

// More on github.com/Dangalcan ↳ View all repositories

// 03 · Toolbox

The tools I
reach for first.

From public repos
and current work

// Languages

  • TypeScript
  • JavaScript
  • Python
  • Bash
  • SQL
  • HTML & CSS

// Frameworks & runtimes

  • Node.js
  • Express
  • React
  • Electron
  • Flask
  • pandas
  • Vitest
  • pytest
  • Playwright
  • Selenium

// Infra & tooling

  • Docker
  • Kubernetes
  • Slurm
  • vLLM
  • LiteLLM
  • Nginx
  • Caddy
  • GitHub Actions
  • Linux
  • MongoDB
  • Redis
  • SQLite

// Operating principles

  1. Fewer dependencies, more self-hosting.

    Each dependency is a promise someone else has to keep. Prefer what ships with the platform, and host the rest yourself.

  2. No numbers nobody measured.

    A claim like twice as fast needs the experiment behind it. Without one, describe what the system does instead of the improvement.

  3. Optimize for the next reader.

    Including (especially) future me, three months from now, who has forgotten everything and just wants to ship one fix.

  4. Ship small, ship often.

    Tiny green CIs beat large red ones. A merged hundred-line PR is worth ten thousand-line drafts in a stale branch.

// 04 · Who

A short
self-introduction.

$ whoami

$ whoami
  daniel.galvan, PhD student, software engineer and lifelong
  terminal dweller.

$ cat about.txt
  I build software around service level agreements and the
  infrastructure behind them. I contribute to open source projects
  on the side and treat side projects as a way to keep learning in
  public.

$ cat education.txt
  PhD in Computer Engineering, University of Seville, in progress
  Master in Software Engineering: Cloud, Data Engineering and
    IT Management, University of Seville
  Degree in Software Engineering, University of Seville

$ cat experience.txt
  now   EOSC-ARENA, Horizon Europe project on generative and agentic
        AI for research. Service level agreements for LLM services
        and agents.
  now   LEIA, AI assistants for software engineering training.
  now   ALMA, the LLM-as-a-service platform of the University of
        Seville. Lead developer and architect, maintainer and in
        charge of its deployments.
  2026  CERN openlab summer student, Geneva. Git-based declarative
        configuration for the CERN Tape Archive (CTA) catalogue.
        1st in the Lunar Lander RL challenge and 2nd in the LHCb
        cooling challenge.
  2026  sla-wizard. Implemented its plugin system and developed its
        plugins, which configure the gateways of ALMA.
  2025  Lucid Chain, final degree project. SLA monitoring for chains
        of services with Governify. Published at JCIS 2025 and
        awarded best Software Engineering final degree project at
        the ETSII.

$ ls ~/interests/
  sla-governance/     service-chains/      llm-infrastructure/
  kubernetes/         developer-tooling/   ai-in-education/

$ cat availability.txt
  Open to research collaborations on SLAs, LLM infrastructure and
  AI agents. Remote friendly. Spanish time zone (CET).

$ 

// 05 · Contact

Let's
build something
good.

Start a conversation