Model
GuaranIA-Gemma
Open-weight LLM based on Google Gemma-4 and adapted for Guarani tasks. To be publicly available on HuggingFace soon.
TBD
Bringing the Guarani into the digital world through AI
GuaranIA
GuaranIA is a project funded by the IDB Lab and developed by CIDIT to advance digital inclusion for Guarani-speaking communities through open AI infrastructure, including models, datasets, tools, and applications that improve access to digital products and services.
Technical Contributions
LLMs, Datasets, and Tools for Guarani.
Project's technical outcomes for building AI infrastructure in low-resource language settings.
Models
In progress
Model
GuaranIA-Llama
Open-weight LLM based on META Llama-3.1 and adapted for Guarani tasks. To be publicly available on HuggingFace soon.
TBD
Datasets
Corpus
Kuatia
Curated collection of more than 40 real and synthetic sources of Guarani documents, including articles, websites, books, and social media content.
CC-BY-4.0View on Hugging Face
Benchmark
MMLU (Guarani)
Professionally translated version of the MMLU Lite benchmark for Guarani. To be publicly available on HuggingFace soon.
CC-BY-4.0
Benchmark
MGSM (Guarani)
Participatory translated version of the Multilingual Grade School Math (MGSM) benchmark for Guarani. To be publicly available on HuggingFace soon.
TBD
Benchmark
WLNI (Guarani)
Participatory translated version of the Winograd benchmark for Guarani. To be publicly available on HuggingFace soon.
TBD
Tools & Apps
Package
GuaraScrapper
Web crawler designed to traverse public websites and collect textual content written in Guarani.
Apache-2.0View on GitHub
Package
GuaraTune
Framework for adapting open-weight models to Guarani through CPT and SFT on Guarani corpora.
GNU GPLv3View on GitHub
Web Application
Mboehara Digital
Platform to facilitate the execution of participatory translation and annotation initiatives.
TBDView online
Mobile Application
AIContigo
AI-powered WhatsApp-based app to help patients follow their medical treatments.
TBD
Methodologies
Methodology
Corpus Pipeline
Methodology for creating Kuatia, the Guarani corpus, through state-of-the-art pre-processing techniques and language-experts validation.
TBD
Publications
Research, reports, and project writing.
Public articles, papers, release notes, and technical reports produced throughout the project.
Medium article
How do modern AI tools understand Guarani?
A blog article introducing experiments to study how state-of-the-art automatic speech recognition (ASR) tools handle Guarani, comparing the performance of five models, including META’s Omnilingual ASR.
Medium article
How well do today’s AI models handle Guarani?
A blog article introducing a comparative study that evaluates 14 state-of-the-art generative AI models on Guarani translation tasks, revealing a clear trade-off between translation fidelity and linguistic validity.
GuaranIA team
[some of the] people making GuaranIA possible.
Luca Cernuzzi
Project Coordinator
Software engineering professor and senior researcher with doctorates from Italian universities, 130+ publications, and extensive R&D, consulting, and public-sector advisory experience.
Jorge Saldivar
Senior AI Research Engineer
Computer Scientist with a Ph.D. in Information and Communication Technologies from the University of Trento, Italy and experience in AI, NLP, Data Science, and Computational Social Science.
Marvin Aguero
Senior AI Research Engineer
NLP engineering researcher with a PhD in Computer Science from the University of Granada, Spain, and a specialization in text mining, machine learning, and data processing for low-resource languages.
Teresa Gamarra
Community Coordinator
Senior risk, communication, and social policy expert with 35+ years leading humanitarian, public-sector, research, and international cooperation projects.
Luis Chiruzzo
AI/NLP Technical Collaborator
Assistant professor at Universidad de la República (UdelaR), Uruguay, with Computer Science Engineering training and MSc and PhD degrees from UdelaR. Expert in NLP for low-resource languages
Agustin Lucas
AI Research Engineer
Machine Learning Engineer specializing in production GenAI, NLP, RAG, multimodal systems, low-resource translation, and cloud-based ML infrastructure.
Andrea Báez
AI Research Engineer
Computer Scientist and Information Engineer with expertise in AI, machine learning and data science.
Cecilia González
Software Developer
Computer engineering student and Python developer experienced in REST APIs and NLP projects.
Carlos Lugo
Software Engineer
Software Engineer with experience in backend and frontend web development, and machine learning.
Raquel Insfrán
Software Developer
Computer Engineering student and certified MERN developer skilled in web applications, databases, teamwork, creative problem-solving, and continuous learning.
Ramón Araujo
Software Developer
Computer Engineering student with expertise in software development, data processing and analysis.
Founders and partners
