Building
resilient
distributed LLMs.
I'm Miquel — a PhD student at Northeastern working on large language models that survive node failures, run on the edge, and stay accurate under real-world constraints.
I work at the intersection of machine learning and distributed systems. Models are getting larger; the devices that need to run them aren't. My research is about closing that gap — keeping inference fast, reliable, and accurate when the compute is spread across many nodes that can fail at any moment.
What I'm working on.
Three threads, all aimed at making large models practical when compute is distributed, unreliable, or sitting on devices at the edge.
Resilience for Distributed LLMs
Extending JARVIS, a framework that splits LLM layers across edge devices. We add peer-to-peer recovery and layer redundancy so the system keeps serving when nodes fail. Tested on Gemma-2B across 18 software-defined radios in the NSF Colosseum emulator and a 7-node Raspberry Pi cluster.
Communication-aware DNN pruning
Training networks for distributed deployment so they need less inter-device communication while staying accurate. Pruning and placing neurons in CNNs and MLPs so they run well on the edge under mixed network conditions.
Knowledge editing in LLMs
Looking at more efficient ways to update or correct specific facts and behaviors inside large language models — without full retraining, and without breaking everything else the model knows.
The path here.
From data science in Barcelona to distributed LLMs in Boston.
PhD, Computer Engineering
Northeastern University · Boston
Distributed LLMs, edge AI, and resilient inference. Advised in the Genesys lab.
Undergraduate Research Assistant
Genesys Lab, Northeastern · Boston
Bachelor's thesis on AI for wireless — signal classification with deep models. Hooked me on AI + wireless and started my path here.
AI Engineer Intern
Abi Global Health · Remote
MLOps monitoring and ML model work for a global tele-health platform.
BSc, Data Science & Engineering
Universitat Politècnica de Catalunya · Barcelona
GPA 8.89/10. Honors in Discrete Math & Logic, Mathematical Optimization, Search & Analysis of Information, and Advanced Topics of Data Science.
Let's build something interesting.
Always happy to talk about distributed ML, edge inference, or how to make models survive the real world.