Software Library
The BSC AI Factory provides a curated software library of ready-to-use tools, frameworks and AI models, fully integrated into the MareNostrum5 environment and aligned with our sector-specific expertise. These resources are designed to help users accelerate development, reduce setup time, and scale solutions efficiently across domains. Whether you are experimenting with language models, training scientific simulators or deploying analytics pipelines, our library offers direct access to high-impact technologies tailored to your needs.
Common Services
-
Salamandra 2B & 7B: Multilingual Large Language Models developed at BSC, available for general-purpose reasoning and text generation in multiple European languages.
-
ALIA 40B: High-capacity multilingual model built for advanced tasks in reasoning, summarisation, and content creation.
-
Embeddings Models: Versatile vector models optimised for semantic search, recommendation systems, and similarity analysis.
-
LangTechLab Models: A collection of language-focused models designed for fine-tuning, prompt engineering and alignment in multilingual contexts.
-
Egida: A framework for evaluating the safety and robustness of LLMs using adversarial prompting and stress tests.
Health, pharma and biotech
-
ALOE: Medical-grade LLM trained for clinical question answering and evidence-based reasoning.
-
Prompt Engine: Custom prompt engineering interface to test, optimise and evaluate prompts for medical AI applications.
Agriculture, climate and blue economy
-
FDB, EarthKit, ANEMOI, Zarr: Libraries and tools for processing and analysing geophysical and environmental datasets.
-
ANEMOI, eMONARCH, AIFS…: HPC-ready frameworks for training and deploying AI models in climate, air quality, and agriculture.
-
Common Repository of BSC LLMs: Fine-tuned language models for environmental and scientific applications, adapted for specialised queries.
-
All the above: Integrated support for selection, preparation and use of environmental data across use cases.
Energy
- SOD2D: Solver for 2D dynamics and surrogate models, enabling faster design and optimisation loops.
-
pyLOM: Lightweight ocean model for rapid simulations in energy and environmental scenarios.
-
ALYA: Multiphysics simulation code for combustion, fluid dynamics, and solid mechanics.
Communications and media
-
NagareDB: Real-time data management platform optimised for NLP and media streaming workflows.
-
SuSy: AI-based tool to detect synthetic images and identify the generative model behind them.
Public sector
- Common Repository of BSC LLMs: Adapted models for public policy analysis, simulations and decision support.
-
ANEMOI, eMONARCH, AIFS…: Infrastructure for real-time AI forecasting in climate, mobility, and environmental planning.
Take your start-up, SMEs and public
bodies to the next level
Contact our One-Stop Shop Manager to get started, your AI journey begins here.