key points
Numendo designed, built, and optimized a cutting-edge AI cloud platform for the EBU, providing flexibility and robustness for the rapid integration of innovative applications. Leveraging an agile approach, advanced technologies (Docker, AWS, Terraform, LLM), and rigorous optimization, this platform strengthens their technological innovation capacity in Europe.
3
key intervention phases led by an expert squad
2
cutting-edge technologies integrated with RAG methods
CONTEXT
The European Broadcasting Union (EBU) sought to develop a flexible cloud platform capable of integrating innovative artificial intelligence applications. The objective was to establish an agile and high-performing infrastructure to host, test, and rapidly deploy cutting-edge AI solutions, such as video facial recognition and advanced natural language processing (NLP).
intervention
To address this complex requirement, Numendo intervened across three key phases:
• Design & Discovery Phase:
○ Facilitation of collaborative workshops to design the technical and functional architecture tailored to EBU’s requirements.
○ Complete UX/UI design on Figma to define expected application behaviors.
○ Definition of a target architecture relying on LLM models (LLama 2 and LLama 3), incorporating advanced methods such as Retrieval-Augmented Generation (RAG).
• Implementation Phase:
○ Complete technical setup management, including AWS infrastructure, front-end, back-end, and data management.
○ Infrastructure management via Terraform for dynamic scaling under load spikes.
○ Encapsulation of AI components into Docker containers using FastAPI and Python for streamlined deployment.
○ Deployment of a centralized authentication layer secured by Keycloak.
○ Training and fine-tuning of dedicated AI models for video facial recognition and advanced NLP.
• Optimization Phase:
○Thorough AWS resource optimization to reduce operational costs through intelligent compute instance management.
○Smooth transition from Proof of Concept (PoC) to a reliable, high-performing operational infrastructure using a DevOps approach and robust software architecture.
○Ongoing agile squad support combining data science and containerization experts to ensure optimal AI application management.
• Design & Discovery Phase:
○ Facilitation of collaborative workshops to design the technical and functional architecture tailored to EBU’s requirements.
○ Complete UX/UI design on Figma to define expected application behaviors.
○ Definition of a target architecture relying on LLM models (LLama 2 and LLama 3), incorporating advanced methods such as Retrieval-Augmented Generation (RAG).
• Implementation Phase:
○ Complete technical setup management, including AWS infrastructure, front-end, back-end, and data management.
○ Infrastructure management via Terraform for dynamic scaling under load spikes.
○ Encapsulation of AI components into Docker containers using FastAPI and Python for streamlined deployment.
○ Deployment of a centralized authentication layer secured by Keycloak.
○ Training and fine-tuning of dedicated AI models for video facial recognition and advanced NLP.
• Optimization Phase:
○Thorough AWS resource optimization to reduce operational costs through intelligent compute instance management.
○Smooth transition from Proof of Concept (PoC) to a reliable, high-performing operational infrastructure using a DevOps approach and robust software architecture.
○Ongoing agile squad support combining data science and containerization experts to ensure optimal AI application management.
results
• Rapid launch of a powerful, scalable AI cloud platform capable of efficiently integrating new innovative applications.
• Production deployment of high-performing AI solutions, including facial recognition and semantic analysis, delivering precise and reliable results.
• Controlled operating costs achieved through thorough cloud infrastructure optimization.
• Successful support for a major European media player in realizing its technological ambitions and strengthening its innovative market positioning.
• Production deployment of high-performing AI solutions, including facial recognition and semantic analysis, delivering precise and reliable results.
• Controlled operating costs achieved through thorough cloud infrastructure optimization.
• Successful support for a major European media player in realizing its technological ambitions and strengthening its innovative market positioning.
