As part of its platform-centric and european growth strategy, ARTE faces major strategic challenges driven by the accelerating pace of digital use and technological changes. To support the channel’s key initiatives in data and AI, this Dynamic Purchasing System (DPS) is intended to pre-qualify partners capable of providing specialized services, delivering high-performance infrastructure, or dedicated software solutions. This new DPS will provide an agile environment for rolling out and testing such solutions, enabling us to anticipate future developments, explore new opportunities and continually adjust our priorities to emerging needs through an iterative approach that integrates the latest technological advances. The purpose of this dynamic purchasing system (DPS) is to cover the needs of ARTE G.E.I.E. (Strasbourg) and ARTE FRANCE (Issy-les-Moulineaux). A purchasing consortium has been set up for this purpose. ARTE G.E.I.E. is the coordinator of this consortium.

LOT-0001
2512501
Category 1: Data Architecture.
• Design, implementation, and monitoring of robust and scalable data architectures and models • Coordination of collaboration with technical teams to ensure architectural integrity and coherence • Evaluation, selection, and optimization of technologies and tools for data infrastructure, particularly Azure Data Factory or equivalent solutions • Formalization and deployment of best-practice references for data management • Diagnosis and resolution of complex technical issues related to data architectures

LOT-0002
2512502
Category 2: Data Engineering and Security.
• Development of a scalable and high-performance infrastructure for large data sets • Management of data ingestion processes, processing, and distribution • Establishment and strengthening of connectors to third-party services (middleware, APIs) • Proactive detection and remediation of vulnerabilities (penetration tests, security audits, threat monitoring) • Ensuring compliance with data protection regulations (GDPR and associated standards) throughout the data lifecycle • Implementation of advanced protection methods (encryption, anonymization, IAM) • Strengthening architectures to prevent cyberattacks and to comply with applicable regulations • Creation, structuring, and updating of technical documentation to facilitate knowledge transfer

LOT-0003
2512503
Category 3: Data Analysis.
• Collection and organization of data while ensuring data quality in close collaboration with editorial, technical, and legal departments • Performance analysis using KPIs and development of appropriate improvement proposals • Design and automation of dashboards and reports • Presenting complex analytical results in an understandable manner for non-technical audiences • Monitoring analytical processes to ensure compliance and reliability • Ensuring compliance with data protection regulations (GDPR, "Privacy by Design" and "Privacy by Default") in all analysis workflows • Creation, structuring, and updating of technical documentation to facilitate knowledge transfer • Use of available BI and data visualization tools to create reports and dashboards for internal stakeholders • Providing studies and strategic recommendations (tagging, segmentation, marketing attribution, scoring, evaluations of data quality, user behavior, etc.)

LOT-0004
2512504
Category 4: Data Management and Governance.
• Extensive experience in data lifecycle management and use of data management tools • Management or enrichment of synthetic datasets for test runs or to fill data gaps • Annotation and classification of multimodal datasets - text, images, audio, video - for training AI models and enriching knowledge databases • Ensuring compliance with data protection regulations (GDPR, AI Regulation) throughout the data lifecycle • Conducting audits and continuous monitoring of data quality and security • Monitoring compliance as well as metadata concerning transparency and traceability

LOT-0005
2512505
Category 5: Data Science (Applied Data Sciences).
• Analysis and evaluation of structured and unstructured data to develop predictive, descriptive, and prescriptive models • Creation and optimization of AI algorithms for business needs (recommendation, classification, segmentation, generation, multimodal analysis) • Collaboration with technical and subject matter teams to integrate innovative approaches to combine various AI techniques • Translating AI solutions into high-performance, scalable, and maintainable services and APIs • Ensuring compliance with data protection and AI-related regulations (GDPR, AI Regulation) and integrating the principles of fairness, explainability, and ethical AI throughout the model's lifecycle • Monitoring model performance, drift, and data quality in production; iterative retraining of models as needed and their updating

LOT-0006
2512506
Category 6: Machine Learning Engineer and Specialists for Multimodal LLM.
• Identification, adaptation, and fine-tuning of open-source models (vision, audio, language, text) for ARTE-specific audiovisual use cases • Implementation of advanced interaction and optimization strategies - prompt engineering, retrieval-augmented generation, LoRA/distillation, quantization, pruning - to maximize accuracy, latency, and editorial relevance • Architecture of multimodal end-to-end pipelines for merging video, audio, images, and text, enriching metadata, optimizing indexing, and automating the production workflow • Evaluation, comparison, and continuous improvement of AI systems against both technical and business process-oriented KPIs • Designing and implementing multimodal cross-data and cross-modality solutions to enhance the predictive and generative capabilities of the models • Management and optimization of the technical infrastructure required for deploying, monitoring, and adjusting the models in production (GPU/TPU clusters, inference gateways, scaling policies) • Coordination of the integration of AI services into existing media systems (MAM/PAM, NRCS, transcription and distribution pipelines, software information systems) in collaboration with product owners and tech leads • Structuring, enriching, and validating audiovisual datasets according to editorial, cultural, and ethical standards • Monitoring, evaluating, and continuously improving the models under real conditions (accuracy, explainability, cultural relevance, CO2 footprint) • Coordinating the alignment of AI solutions with strategic goals as well as the legal framework together with multidisciplinary teams (GDPR, AI Regulation) • Drafting and maintaining comprehensive technical documentation to ensure knowledge transfer

LOT-0007
2512507
Category 7: UX/UI Design with a Specialization in Data and AI Applications.
• Designing intuitive user interfaces for AI tools and analytical dashboards to maximize the value of predictive and generative models • Creating advanced data visualization systems to transform multidimensional information into clear and actionable insights • Developing user-friendly interfaces and interactions to promote transparency and comprehensibility of AI models (e.g., through prompting, explanatory visualizations, feedback loops) • Conducting targeted usability tests to assess user understanding, trust, and overall experience with AI systems • Defining and applying ethical design principles to ensure transparency, explainability, and responsible use of AI • Ensuring compliance with accessibility standards (WCAG / RGAA) and data protection regulations (GDPR, AI Regulation) throughout the design process • Producing, structuring, and updating design documentation (design system guidelines, interaction models, user research reports) for more efficient collaboration and better knowledge transfer

LOT-0008
2512508
Category 8: AI Strategy and Expertise for Newsroom, Production, and Editorial.
• Strategic consulting at the intersection of AI, data, and the specific media workflows (editorial, content production, post-production) • Implementing editorial, journalistic, and production requirements into data- and AI-based solutions • Identifying and prioritizing high-quality use cases (storytelling support, optimizing editorial workflows, production planning, enhancing metadata, cost tracking, etc.) • Developing AI and data strategies that specifically support the operational goals of editorial, journalism, and production within a public media environment • Assisting in organizational change, transformation processes, and capacity building for the long-term integration of AI solutions • Evaluating the impact of AI on journalistic workflows, editorial values, and public service missions; defining success KPIs and ROI metrics • Advising on the implementation of AI in editorial and production core systems – from planning through editing workflows to archiving and newsroom platforms • Ensuring compliance with ethical and regulatory standards (AI Regulation, GDPR, copyright, editorial responsibility, diversity, and inclusion) • Creating, structuring, and updating strategy papers, roadmaps, and best-practice guidelines to facilitate effective knowledge transfer

LOT-0009
2512509
Category 9: Strategic Audit and Consulting.
• Evaluation of the performance, quality, and compliance of AI and data-driven systems across all areas of ARTE (production, editorial, distribution, support)
• Comparison of proprietary solutions with open-source and SaaS alternatives, highlighting technical, economic, and ecological trade-offs
• Presentation of actionable recommendations for cost optimization, CO2 footprint reduction, scalability, security, and maintainability
• Identification and quantification of project risks (security, compliance, ethics, editorial integrity, copyright) and proposing risk mitigation plans
• Guidance on ongoing alignment with relevant regulations and standards (AI Act, GDPR, copyright, accessibility, obligations as a public broadcaster)
• Ensuring strategic and technological oversight to identify emerging opportunities or threats in the fields of AI, data, and media innovation
• Strengthening internal capacities through workshops, training events, and structured knowledge transfer programs
• Co-designing roadmaps and governance frameworks for embedding AI while managing change and involving stakeholders
• Preparation, structuring, and updating of audit reports, maturity assessments, and decision dashboards to support transparent and data-driven governance

LOT-0010
2512510
Category 10: Product and project management specializing in AI and data.
• Analysis and translation of business requirements into functional specifications for data- and AI-driven solutions
• Development and management of product roadmaps in line with ARTE's innovation strategy and public remit
• Identifying, analyzing, and strategically prioritizing innovation potentials in the fields of data, artificial intelligence, and new technologies
• Coordination of collaboration between internal stakeholders and external partners to maximize value creation and mitigate risks
• Promotion and comprehensive advocacy for data processing and AI-based initiatives to internal and external audiences (demos, showcases, success stories)
• Guidance on organizational transformation and change management for faster deployment of data-centered solutions
• Alignment with interoperability standards and applicable regulations (GDPR, AI Regulation, copyright)
• Production, structuring, and updating of product artifacts and project documentation (backlogs, OKRs, risk registers, lessons learned) to support transparent governance

LOT-0011
2512511
Category 11: Strategy development in open source and technology governance.
• Developing and tracking a comprehensive open-source strategy tailored to ARTE's mandate, its technical landscape, and the regulatory context
• Establishment of governance models for the use of open source software (contribution workflows, license selection, risk assessment, publication policies)
• Legal and organizational advice on licenses (MIT, GPL, AGPL, Creative Commons), supplier agreements, and special cases related to AI (model weights, datasets, prompts)
• Designing sustainable processes for managing and publishing OSS projects (documentation, versioning, disclosure of security measures, metrics for 'Project Health Check')
• Establishing community management practices - communication channels, integration paths, moderation policies, and external visibility
• Advising on strategic partnerships with OSS foundations, public broadcasters, research consortia, and EU initiatives
• Monitoring legal, technological, and societal trends to anticipate potential impacts on ARTE's AI and data systems and ensure ongoing compliance with regulations (AI Regulation, GDPR)
• Creation, structuring, and updating of corporate policy documents, strategic presentations, and training materials to promote internal dissemination and knowledge transfer

LOT-0013
2512513
Category 13: Co-development environments.
Solutions for engineering teams for efficient design, implementation, testing, and collaboration utilizing the latest AI advances
• Integrated collaborative platforms: Enterprise-grade solutions for real-time synchronized development environments with version management optimized for AI/ML projects
• AI-enhanced development support systems: Tool suites for significantly increasing the productivity of engineering teams through contextual intelligent autocompletion, assisted program generation, and automated refactoring for data science environments
• Advanced technologies to support self-generated documentation, intelligent debugging, and algorithmic optimization to significantly shorten development cycles and accelerate production onboarding

LOT-0014
2512514
Category 14: Platforms for model access and orchestration.
Technological solutions for easier discovery, evaluation, and integration of AI models while supporting emerging architectures based on autonomous agents
• Unified platforms for model access: centralized environments for structured catalogs of pre-trained models with standardized interfaces and multi-vendor compatibility to abstract the underlying access layers
• Compute and energy-efficient model architectures: optimized implementations with low energy and compute requirements, designed for resource-constrained deployments and edge computing applications
• Verticalized ecosystems per field and modality: collections of models specifically tailored to priority business verticals (especially audiovisual media), covering the entire modality spectrum (text, image, audio, multimodal)
• Enhanced frameworks for model orchestration: automated decision systems capable of dynamically routing or assembling models according to parameterizable constraints regarding latency, cost, accuracy, and application context
• Agentic architecture tool suites that help engineering teams design, parameterize, and monitor autonomous or semi-autonomous LLM agents that link predictive models and external services in highly automated end-to-end workflows (capability for planning, executing third-party tools, and sequential thinking)

LOT-0015
2512515
Category 15: Lifecycle management of AI models.
Integrated technical solutions for complete lifecycle management of AI models - from development to production
• Deployment options for multiple environments: platforms ensuring coherent and seamless deployment of model artifacts across heterogeneous environments (public/private cloud, edge computing, on-premise infrastructures), with extended versioning and rollback features
• Intelligent autoscaling and dynamic resource optimization: sophisticated systems for automatically adjusting IT infrastructure to changing loads with optimal throughput while achieving cost efficiency
• AI-specialized CI/CD pipelines: end-to-end orchestration from training to controlled deployment, including automated tests, multidimensional validations tailored to the specifics of predictive models, and continuous delivery
• Infrastructure for lifecycle management and security: IT systems for progressive deployment (canary, blue-green), automated rollback for anomaly detection, granular access controls, and real-time monitoring for risk mitigation and compliance

LOT-0016
2512516
Category 16: Databases and Retrieval-Augmented Generation (RAG).
Highly advanced solutions for data structuring and semantic search within AI architectures
• Semantic and vector-based data platforms: semantic and vector-based search engines with clustering, embedding, and multimodal retrieval features for media or editorial purposes
• Specialized database infrastructures: highly available, fault-tolerant, and optimized storage solutions for complex analytical workloads or distributed vector algebra (SQL, NoSQL, vector, graph)
• Data ingestion and transformation: automation frameworks for capturing, transforming, and indexing data from multiple sources and in multiple formats, with complete pipelines for semantic enrichment or modality-crossing vectorization (text, image, audio, video)
• Contextual document access systems: solutions for accurate document retrieval through enhanced contextual understanding, intent-aware analysis while maintaining semantic links between entities
• Verticalized RAG architectures: specific RAG frameworks with continuous learning capabilities, incremental updates without full re-indexing, and sophisticated memory management strategies

LOT-0017
2512517
Category 17: Monitoring, prompt engineering, and quality assurance. Solutions for monitoring, control, performance optimization, and reliability of AI models in production environments
• Platforms for observability and advanced monitoring: Integrated solutions with real-time dashboards and configurable alerts for continuous monitoring of usage and performance metrics with anomaly detection and predictive analysis of model behavior deviations
• Infrastructures for systematic evaluation and benchmarking: Specialized frameworks for continuous evaluation of predictive models and RAG systems based on specific expert criteria for automated benchmarking for predecessor versions and industry standards
• Mechanisms for quality assurance and mitigating hallucinations: Techniques for proactive identification and containment of hallucinations, improving factual coherence, and validating the quality of content generated across multiple parameters
• Provisions for compliance and technical governance: Automated platforms for assessing and certifying compliance with regulations (GDPR, AI regulation) and adherence to audit standards for algorithmic explainability, secure logging, and granular traceability of access and interactions

LOT-0018
2512518
Category 18: AI Model Training and Fine-Tuning.
Solutions for optimizing training processes, fine-tuning, and managing learning corpora
• Integrated platforms for supervised fine-tuning: Specialized environments for adapting foundation models to targeted use cases while optimizing the required data volumes Comprehensive support for supervised learning, instruction tuning, and RLHF (Reinforcement Learning from Human Feedback)
• End-to-end automation of training pipelines: Complete orchestration of the training cycle with enhanced automation of hyperparameter tuning, distributed parallelization, and reproducible test environments
• Central management of training corpora: Unified platforms for capturing, collaborative annotation, and semantic enrichment of datasets including quality assurance and traceability
• Optimization of computing power and efficiency: Solutions to reduce CO2 footprint and training costs through LoRA, QLoRA, quantization, and pruning while maintaining the predictive performance of models
• Strategies for data enrichment: Systematic methods for enriching or diversifying training datasets to improve the robustness and generalizability of models at minimal additional costs
• Continuous evaluation and monitoring of models: Integration of performance and drift metrics with real-time dashboards to ensure quality, reliability, and compliance throughout the model lifecycle

LOT-0019
2512519
Category 19: Multimodal Processing, Translation & Content Generation.
Highly advanced solutions for automated processing, intelligent annotation, and AI-supported generation of multimodal content and enriched metadata
• Automatic speaker recognition and subtitling: Highly accurate ASR systems (Automatic Speech Recognition) for specialized fields as well as a variety of accents with multilingual alignment technology for image-to-image synchronization of transcriptions with audio/video streams
• Machine translation and localization: Customizable machine translation for subtitles, transcripts, and metadata with integrated glossary management, adaptation to specific disciplines, and metrics for quality assessment
• Software suite for post-production or editorial workflows: Modern subtitle editors with automatic speaker attribution, stylistic unification, and timing management; hybrid approaches combining automated processing and targeted manual post-editing
• Extraction and enrichment of multimodal metadata: for semantically identifying and classifying key elements (people, objects, places, actions, topics) in media content, as well as generating specialized AI frameworks focused on industry-standard metadata
• Collaborative video annotation tools: Collaborative work environments to support semi-supervised annotation of audiovisual content with AI-assisted segmentation, object tracking, and context-aware detection of visual elements