I work on the infrastructure that trains and serves models: GPUs, clusters, storage, and the platform on top. Currently building it from day one at Speridlabs, a spatial AI lab working on foundation models for the 3D world.
| Role | Where | When |
|---|---|---|
| Founding AI Infrastructure Engineer | Speridlabs | 2026 – present |
| AI Research Assistant | Telefónica Research | 2026 |
| MLOps Engineer | Multiverse Computing | 2026 |
| AI Engineer Intern | CERN | 2025 |
| LLVM Software Engineer | Barcelona Supercomputing Center | 2024 – 2025 |
| Data Scientist | Mango | 2023 – 2024 |
B.Sc. in Computer Science at UPC. International double M.Sc. in Data Science at UPC and the Università degli Studi di Padova, with a thesis on the price of advertising inside LLM conversations.
- AI infrastructure at Speridlabs. Train, serve, repeat, cheaper every time.
- Troiani. A sub-billion language model, from tokenizer to eval, from scratch.
- The Scaling Journey. Public notes on LLMs, world models, AI infrastructure, and the harness around models.
Models. LLMs, world models, training, evaluation.
Infrastructure. Kubernetes, GPU scheduling, distributed training, inference serving, storage.
Harness. Agents, evals, and how people actually experience AI.
- 🏋️ Gym, calisthenics, BJJ, bouldering, marathon running, salsa, piano, and as much nature as I can get.
- 🗣️ Spanish, Catalan, Italian, and English. Learning German and Ukrainian.
- 🎁 Savouring the gift of life.
- ⚡ The indomitable human spirit and its will, will always persevere against the cruel indifference of the world.