PhD in Natural Language Processing (NLP) by University of the Basque Country UPV/EHU, IXA Group, and HiTZ Basque Center for Language Technologies.
During my time at HiTZ, I specialized in data acquisition, training, and evaluation of Large Language Models (LLMs). Notable contributions include: GoLLIE, a 34B-parameter LLM achieving state-of-the-art results in zero-shot Information Extraction; MedMT5, the first open-source multilingual text-to-text model for the medical domain; and Latxa, a Basque instruction-tuned model trained on the Leonardo Supercomputer using up to 512 GPUs and 128 nodes, delivering performance comparable to GPT-4o and Claude Sonnet.
At Krea.ai, I developed a personalized image recommendation engine based on user interaction data and deep learning methods. I have designed and trained SigLIP-style embedding models for artistic style-based image retrieval, and contributed to the development and optimization of codebases for fine-tuning video (Wan 2.1) and image (Flux) models, with a focus on performance improvements and distributed training using techniques such as FSDP2. I also implemented and deployed generation services for 3D models and textures, which power features used by thousands of users on the Krea platform.
Currently an LLM Researcher at Multiverse Computing.
Exponential Fellow 2025 https://www.goexponential.org/directory/fellows/iker-garcia-ferrero
In my free time, I develop veridika.ai, an AI agent framework for real-time fact-checking.
At Krea.ai, I developed a personalized image recommendation engine based on user interaction data and deep learning methods. I have designed and trained SigLIP-style embedding models for artistic style-based image retrieval, and contributed to the development and optimization of codebases for fine-tuning video (Wan 2.1) and image (Flux) models, with a focus on performance improvements and distributed training using techniques such as FSDP2. I also implemented and deployed generation services for 3D models and textures, which power features used by thousands of users on the Krea platform.
Currently an LLM Researcher at Multiverse Computing.
Exponential Fellow 2025 https://www.goexponential.org/directory/fellows/iker-garcia-ferrero
In my free time, I develop veridika.ai, an AI agent framework for real-time fact-checking.
