We are innovation pioneers. We invest in research to provide the most advanced ML and AI solutions to our business clients and research collaborators.
OPTIMINER leverages artificial intelligence, digital twins, and advanced Critical Raw Material extraction technologies to enhance resource efficiency, reduce environmental impact, and improve the sustainability of Europe’s mining sector. The project will develop and demonstrate its innovative solutions across five mining sites in Spain, Greece, Poland, Finland, and Chile.
Our Role: LIBRA AI Technologies leads the development of an AI-enabled Decision Support System (DSS) that integrates real-time data analytics, alerting system, and other project’s intelligent components, to optimize CRM recovery and resource efficiency. Additionally, LIBRA is responsible for the AI-augmented Market Observatory, a tool designed to analyze and forecast CRM market trends, supply chain risks, and economic factors, insights into the market dynamics, supporting strategic decision-making for industry stakeholders.
Duration: Jan 2025 – Dec 2028
Call Identifier: HORIZON-CL4-2024-RESILIENCE-01
![]()
Our Role: LIBRA AI Technologies was selected in the 1st Innovation Scheme Open Call of ENFIELD, an EU-funded Horizon Europe research and innovation project, and developed a generative AI model for single-cell RNA sequencing (scRNA-seq) data in Alzheimer’s disease. We adapted GPT-based transformer models to biological sequence data and trained two model configurations on 96 curated and standardised Alzheimer’s-specific scRNA-seq datasets, with more than 1.7 million single cells, extending the model vocabulary with 2,000 Alzheimer’s-relevant genes. The resulting models identify gene co-expression networks, cellular state transitions and biomarkers associated with the disease. They increased Alzheimer’s vs control classification accuracy from 77% to 83% and Braak stage classification accuracy from 65% to 83%. They also mapped 141 Alzheimer’s-associated gene clusters covering 550 genes, with region-specific activity strongest in the entorhinal cortex and hippocampus, two regions central to disease progression. The results were presented at the IEEE BIBE 2025 conference.
Status: Completed
Duration: 6 months
Call Identifier: ENFIELD 1st Innovation Scheme Open Call (Horizon Europe)
![]()
Our Role: LIBRA AI Technologies develops AdaptiveSight, an attention-aware decision layer that controls the energy use of digital signage. The system estimates viewer attention directly on the device and adjusts screen brightness, refresh rate and content scheduling, so energy use follows real viewership. Its perception layer reuses our on-device eye-tracking research, which received the Springer Best Paper award at ISVC 2025. All processing runs on an edge device next to the screen, and no one is identified or recorded. We validate the system in our computer vision lab on a commercial signage panel fitted with a power meter, and through digital-twin simulations fed by public footfall data. The project targets a 30 to 50% reduction in screen energy use with no loss of attention-validated impressions.
Status: Ongoing. Final report and demonstration due by the end of December 2026
Call Identifier: ELIAS Open Call #2 (Horizon Europe)
![]()
Our Role: LIBRA AI Technologies develops Propensio, an AI lead intelligence and routing engine. Propensio scores each inbound enquiry, whether it arrives by web form, email, chat or API, by its probability of converting, and returns a priority tier and a routing decision within seconds. Each score combines more than 40 behavioural, contextual and intent-text features in a gradient-boosted model, and staff can open an explanation of why an enquiry scored as it did. Propensio is in production at a leading online travel agency, where the top 6% of scored traffic converts at 28%, against a 5% conversion rate before deployment. Propensio was selected for Stage 2 of the AI-on-Demand (AIoD) open call, and we are now turning it into a product that any business can adopt, with scores returned over an API and integration through the tools customers already use.
Status: Ongoing
Call Identifier: AI-on-Demand (AIoD) DeployAI Open Call for the private sector, Stage 2
![]()
Our Role: LIBRA AI Technologies won Challenge 6 of the COSMIC 1st Open Call as the sole beneficiary and is developing SmartTherm, a vision-only thermal comfort recognition system with adaptive cooling control. SmartTherm reads thermal comfort cues, such as skin temperature signatures, fanning and clothing adjustments, from infrared and RGB cameras, with no wearables and no personal identification. CNN models recognise these cues on an edge device at the site, and a reinforcement learning controller adjusts cooling to the comfort state it detects. The system connects through open APIs to the COSMIC data and energy platforms and is being piloted at the Cartuja Urban Lab in Seville. The project targets 20–40% lower cooling energy in public spaces compared with static comfort models, while keeping or improving occupant comfort.
Status: Ongoing, within the 10-month COSMIC support programme
Duration: April 2026 – January 2027
Call Identifier: COSMIC 1st Open Call, Challenge 6 (Horizon Europe)
![]()
Our Role: LIBRA AI Technologies designed and built MOUSSE (Metadata fOcUsed Semantic Search Engine) end to end, with the Institute of Communication and Computer Systems (ICCS) providing Earth Observation expertise. MOUSSE is an LLM-powered semantic search engine for large data portals, demonstrated on marine and Earth Observation data. A fine-tuned large language model extracts the time and location filters from each natural-language query, and a hybrid search on PostgreSQL combines spatial, temporal and semantic vector search, so results match the meaning of the query and not only its keywords. Large result sets are grouped into themed clusters with LLM-generated titles, and users can draw a polygon on a map to search a precise area. MOUSSE builds on our earlier cognitive search work for the GEOSS portal in the EIFFEL project, and its code is released as open source under the MIT licence.
Status: Completed
Duration: June 2024 – May 2025 (12 months)
Call Identifier: NGI Search Open Call #4 (Horizon Europe)
![]()
Our Role: LIBRA leads the respective work related to the development of the CHOICE immersive enablers that will be widely used to facilitate behaviour change and drive action towards food mitigation measures while being responsible for the creation of engaging data storytelling experiences in order to help citizens and industrial actors retain information and see themselves in the story of climate change.
Additionally, the company will perform Randomized Control Trials in the five pilot sites of the project to optimise the design and effectiveness of CHOICE engagement and "green marketing" campaigns. Serving as the project's communication manager, LIBRA will also establish and manage the primary communication channels of the project, ensuring seamless collaboration among all stakeholders for the successful implementation of project communication activities.
Duration: Nov 2023–Dec 2026
Call: HORIZON-CL5-2022-D1-02
Our Role: LIBRA performs Machine Learning based measurement and verification activities to model buildings' energy footprints. The aim is to predict their future energy consumption and offer an independent assessment of the effectiveness of the interventions that PRELUDE partners will apply in the project.
Our approach will be evaluated in 8 different building sites of several sizes and functionalities in four European countries, Switzerland, Italy, Poland, Denmark, and Greece.
Duration: Dec 2020 – May 2024
Call Identifier: H2020-NMBP-ST-IND-2020-singlestage
This project has received funding from the European Union’s Horizon 2020 Research and Innovation Programme under GA N° 958345.
Our Role: LIBRA develops the AI-Enabled business intelligence system of the project that integrates a range of advanced predictive modeling and optimization tools developed by the whole consortium and provides insights and analytics of all the critical mining processes and operations, including productivity, monitoring, employee safety, environmental monitoring, etc.
The system will be deployed in 5 (five) mines.
Duration: May 2020 – April 2024
Call Identifier: SC5-09-2018-2019
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under GA No. 869529.
Our Role: GEOSS is the Global Earth Observation System of Systems that covers many independent Earth observation systems and providers, e.g. satellites, drones, and in-situ sensors, at a global scale. LIBRA is developing the cognitive search engine for GEOSS data based on advanced AI-enabled Natural Language Processing (NLP). This search engine will empower scientists and stakeholders in the Climate Change domain, allowing them to search effectively for data and information using natural language for their queries in a fashion similar to the one we all use in Google search for general inquiries. Moreover, LIBRA will provide tools for the multifaced enrichment and curation of the GEOSS metadata. LIBRA is also the EIFFEL's exploitation manager coordinating the exploitation effort of the project's consortium.
Duration: June 2021 – May 2024
Call Identifier: H2020-LC-CLA-2020-2
This project has received funding from the European Union's Horizon 2020 research and innovation programme under GA No. 101003518.
Our Role: LIBRA develops the project’s Immersive Business Intelligence system - METAMINE – of immersive dashboards with virtual control rooms and a real-time alerting system and leads the Machine Learning Operations (MLOps) for the MASTERMINE ML/AI systems, being responsible, among others, for the design and development of an AI-assisted data mapping framework.
Our approach will be demonstrated in five EU demo cases in Spain, Greece, Finland and Poland and one replication demo in South Africa.
Duration: Dec 2022–Nov 2026
Call: HORIZON-CL4-2022-RESILIENCE-01

Our Role: LIBRA develops an interactive dashboard suite, providing comprehensive views of climate resilience indices, resulting from Nature-Based Solutions (NBS) impact assessments, tailor-made for the CARDIMED stakeholders. The Dashboard suite is part of the CARDIMED NBS interventions management digital platform.
In addition, LIBRA develops engaging data storytelling materials aligned with the project's objectives and prospects for local NBS communities.
Duration: Sept 2023 – Feb 2028
Call Identifier: HORIZON-MISS-2022-CLIMA-01
![]()
Our Role: LIBRA develops a Machine Learning (ML) methodological framework for AI experimentation. This framework includes tools for tasks like time-series forecasting, model training, and explainable AI.
The resulting ML models will enhance project modules such as early warning systems. LIBRA also leads the development of a time-series analysis toolbox, defines ML frameworks, manages data storytelling, and coordinates innovation and exploitation efforts within the project.
Duration: Jan 2024 – Dec 2027
Call Identifier: HORIZON-HLTH-2023-ENVHLTH-02

Our Role: LIBRA's role includes drafting project specifications and developing a gaze prediction model that tracks and predicts viewer focus, enabling real-time video adjustments to improve the viewing experience.
Duration: Jan 2024–Dec 2027
Call: HORIZON-MSCA-2022-SE-01

Our Role: LIBRA AI Technologies serves as the Technical Manager of the HIDDEN project, coordinating cross-partner technical activities and ensuring integration across work packages. LIBRA leads the development of an AI-powered driver state monitoring system that enhances collective perception by capturing driver gaze, alertness, and emotional state using synchronised in-cabin and external cameras. This human feedback layer strengthens automated decision-making in complex urban scenarios and contributes to more ethical, explainable, and human-aligned vehicle behavior.
Duration: July 2025 – June 2028
Call: HORIZON-CL5-2024-D6-01
![]()
