Profile Picture

Oriol Miró López-Feliu

Research Engineer @ BSC AI MSc Student @ UPC, UB, URV

I am a research engineer at Barcelona Supercomputing Center working in the HPAI group. I recently published my first paper (oral presentation!) at CCIA. I am also pursuing a master’s degree in Artificial Intelligence at Universitat Politècnica de Catalunya, Universitat de Barcelona and Universitat Rovira i Virgili. I am also the head of the electronics department of ARM2u, a student-led team deveolping an affordable, user-centered myoelectrical transradial prosthetic arm. Last July, we were invited to the United Nations’ AI for Good Summit to present our project. I am interested in agents, RL, neuroscience, and societal+cognitive effects of AI. In my free time I tinker with robotics, develop projects that improve my quality of life, solve games with AI agents, read, and do sports.

Experience

Barcelona Supercomputing Center

Barcelona, Spain

Member of the HPAI group working on agents, MAS, and RL.

Research Engineer

January 2025 - Present
  • Collaborating on various research projects.
  • Participating in the AI4Europe european project, building the AIoD platform.

ARM2u

Barcelona, Spain

Student-led biomedical engineering team building a affordable myoelectrical transradial prosthetic arm.

Electronics Lead

July 2025 - Present
  • Leading the electronics department.

AI Engineer & EMG Research

January 2025 - June 2025
  • Workin on AI section of prothesis. Focus on electronics, EMG signal decoding, and embedded systems.

T2C

Barcelona, Spain

AI/ML engineering and data projects for clients.

AI Engineer Intern

February 2024 - July 2024
  • Built a GenAI content generator using RAG, web scraping, and LLM tuning (ChatGPT, Claude, Mistral, Llama).
  • Led a video-to-store clothing-retrieval system; delivered an end-to-end pipeline running ~4 FPS.
  • Built a model that scanned newspapers into accessible HTML using OCR and LLMs.

Machine Learning Engineer Intern

July 2023 - September 2023
  • Contributed to time-series forecasting for bottled-water demand; led model development, testing, and optimization.

Education

Bachelor's Degree in Informatics Engineering specialization in Computer Science

Universitat Politècnica de Catalunya - Facultat d'Informàtica de Barcelona

September 2020 - June 2024

240 ects

Average mark: 8.41/10

Exchange studies (Erasmus+ Programme)

The University of Edinburgh - College of Science & Engineering

August 2023 - December 2023

30 ects

Average mark: 9.3/10

Extracurricular Activities

Publications

Listen, Adjust, Act: Adding Communication to Pre-trained Agents via Goal Adjustments

Oriol Miró López-Feliu; Adrian Tormos; Victor Gimenez Abalos

CCIA, 2025

Adds comms to pre-trained goal-conditioned RL by adjusting latent goals, enabling coordination on multi-agent tasks without retraining end-to-end. Tested on the challenging Minecraft environment.

FROM TRAPS TO TREASURE: AN AGENT ENVIRONMENT FOR SPELUNKY 2'S PROCEDURAL DEPTHS

Víctor González Bentué; Oriol Miró López-Feliu; Víctor Gimenez Abalos

Bachelor Thesis (TFG), UPC – FIB, 2025

I co-directed this thesis which introduces spelunkyRL, a customizable Gymnasium environment for Spelunky 2 via a Lua mod, with procedurally generated, partially observable levels that demand long-horizon planning. Supports flexible tasks and efficient parallelism; reveals algorithmic limits—on-policy methods reach ~85% on simplified levels but struggle with hazards and sparse rewards.

Skills

Python

Main language used for AI/ML projects. Used in university, work and personal projects. Also used for scripting and automation.

C, C++

Main languages used in university. Used in low-capacity embedded systems (ARM2u) and other robotics projects.

PyTorch

Primary ML library for research at BSC, and works in university; training all sorts of AI models, from transformers to RL agents.

LangChain, CrewAI

Using agentic frameworks to design tool-using agents and multi-agent crews; planning, memory, RAG tools and evaluators.

LLM Tooling

Running and serving local/hosted LLMs (Ollama); quantization (GGUF), prompt templates, and adapter fine-tuning.

W&B, Hydra

W&B for metrics, artifacts, reports and hyperparameter sweeps; Hydra for structured configs, multirun and reproducible overrides.

HPC

Experience using HPC clusters (such as BSC’s MN5), SLURM, and parallel computing.

Git

Experienced with git-based development. Mostly, use GitHub and Gitlab.

Linux

Using as the main operating system. I’m used to interact with it via the command line. Shell scripting is also one of my strengths.