The team provided updates on the GenAI for Africa project, discussing key work packages, budget allocations, and travel budgets for partners. They reviewed specific work packages focusing on risk research, and the development of GenAI building blocks for smart agriculture in Africa. The group also discussed proposal finalization processes, with work packages responsible for providing content for specific sections and agreeing to skip the September 19 meeting to allow more time for proposal writing.

• All partners: Finalize their work package narratives following the common project narrative
• All partners: Specify more clearly what they will deliver as final products/outcomes of the project
• All partners: Review the budget allocation and provide feedback to SU
• All partners: Focus on addressing multilingual technology and data collection barriers in Africa
• All partners: Identify which tasks they will lead or contribute to in the work packages
• All partners: Decide on the use of the optional 5K budget for travel/dissemination/technical meetings
• STR and NU London: Lead WP1 on analyzing African needs
• Vitalis: Update Work Package 1 description to include the living labs as part of the demand analysis
• Louis: Provide more details about potential collaboration with Deep Learning Indaba and Masakhane communities
• European partners: Identify GenAI solutions that can be provided for the African living labs as requested by Bamba

Serge provided updates on the GenAI for Africa project, highlighting the consortium's focus on three living labs in Africa (UCT, UCAD, and STR) and the addition of Northeastern University as a new partner to replace TU Berlin/FOKUS. The project aims to develop GenAI building blocks targeting smart agriculture in Africa, with a budget of 2 million euros and a 30-month duration. Key work packages were outlined, including analysis of African needs (WP1), identification of EU solutions (WP2), co-construction and customization (WP3), and sustainability/replication. The team emphasized the importance of developing a common narrative highlighting the project's focus on TRL 4-6 technologies and the need to specify the final deliverables more precisely.

The team discussed budget allocations for partners, with Sorbonne and CNR receiving slightly more due to coordination and multiple teams involved, while Northeastern received a lower budget for lighter participation. Travel budgets were set at 15K mandatory for plenary meetings and an optional additional 5K for dissemination or technical meetings, with a maximum total of 20K per partner. The group reviewed Task 1.1-1.4 focusing on risk research and stakeholder surveys, with Vitalis identified as the lead for Task 1.4, though concerns were raised about the need to better integrate living labs into the demand analysis.

The team discussed work package structures and deliverables for a project involving Living Labs. For Work Package 1, they clarified that desk research should focus on specific topics that align with the Living Labs' focus areas rather than conducting broad research, with Stavroula suggesting replacing "desk research" with terms like "gap analysis" in the title. Raffaele presented Work Package 2, which will analyze promising European-centric and open-source foundational models in GenAI, followed by organizing focus groups with experts to identify suitable models and define adaptation guidelines. Panayiotis began explaining Work Package 3, which involves creating a data pipeline and knowledge base from various sources including IoT sensors and agricultural data, though the discussion was cut off at the end of the transcript.

Panayiotis presented a comprehensive plan for developing an agricultural AI system, outlining the creation of user personas, data harmonization, analytics engine, conversational AI agents, and communication layers tailored to farmers and officials. Bamba raised concerns about data availability in African contexts, to which Panayiotis responded that the required data would depend on user requirements and local availability, suggesting environmental measurements, government lists, and best practices as potential data sources. Louis emphasized the importance of incorporating community and endogenous knowledge to mitigate data bias, while Stavroula offered to share the document with their agronomist to provide more specific insights on local agricultural data and indicators.

The team discussed Work Package 4, focusing on living labs planning and deployment. Stavroula presented the structure, which includes preparation, deployment, and assessment tasks. The group agreed to merge tasks 4.1 and 4.2, with EU partners involved in integration and deployment. Raffaele raised questions about the involvement of EU partners in specific tasks, particularly regarding integration between solutions and building blocks. The team also reviewed Work Package 5, which covers the Agri GenAI Africa Kit and replication framework. Stavroula suggested adding tasks for clustering with other projects and policy recommendations. The group discussed the overlap between training in Work Package 5 and Work Package 6, deciding to clarify the distinction between the two. Finally, they agreed to remove the specific task on scientific dissemination, as it would be covered under general communication and dissemination efforts.

The team discussed the process for finalizing a proposal, with work packages responsible for providing content for specific sections including objectives, R&I maturity, and methodology. Stavroula offered to draft Section 2.2 on dissemination, though Serge suggested waiting to avoid duplication. The group agreed to skip the September 19 meeting to allow more time for proposal writing, with a new meeting scheduled for September 26. Raffaele confirmed that all work package content should be provided by Friday, after which the team will merge contributions and create a full draft for review by the end of the following week.