About GAINS

GAINS (Generative AI for Networks and Sustainability) investigates the scientific and technological challenges of building an AI-based automation and decision-support system for telco and cloud operators. By integrating heterogeneous sources (configurations, SLAs, organizational processes, service portfolio, and real-time telemetry), GAINS aims to generate standardized and sustainability-optimized deployment models, including ready-to-use configuration details, technical/procedural documentation, and predictive monitoring and environmental impact insights.

Programme Context

• Programme: RESTART - PNRR / NextGenerationEU
• Spoke: 7 - Green and Smart Environments
• Duration: 8 months
• Technology Readiness Level: from 1 to 4

The Project

GAINS is designed around three integrated operational processes:

• Monitoring and prediction of energy consumption and emissions, including support for environmental compliance audits.
• Optimization of network configurations to reduce errors and implementation time, while improving sustainability.
• Generation of clear, policy-aligned technical documentation to accompany and explain proposed configurations.

The proposed system uses a common control plane to split the overall complexity of the problem into specialized components using an agent-based architecture. The project investigates AI agents implementing Retrieval-Augmented Generation (RAG) with a human-in-the-loop approach to increase reliability, traceability, and operational safety.

Work Packages and Key Tasks

WP1 - Problem Framing & State of the Art (Fundamental Research)

• T1.1: State of the art on LLM agent architectures and energy impact (including RAG and sustainability trade-offs).
• T1.2: State of the art on the energy impact of network and cloud infrastructures (routing, switching, cloud platforms).
• T1.3: Dissemination of preliminary results (publications, events, stakeholder outreach).

WP2 - System Design, Agent Integration & PoC Validation (Industrial Research)

• T2.1: System design (data sources and formats, agent interaction model, interfaces).
• T2.2: End-to-end integration and proof-of-concept evaluation (functional tests, stress tests, constraint adherence).
• T2.3: Dissemination (conferences, workshops, white papers, demos).

WP3 - Agent Prototyping & Evaluation (Industrial Research)

• T3.1: Network Monitor (real-time monitoring and predictive analytics for energy/CO₂/utilization and anomalies).
• T3.2: Network Orchestrator (policy-aware, standardized service deployment for telco/cloud; human-in-the-loop approvals).
• T3.3: Knowledge Manager (automated technical/procedural documentation aligned with policies and SLAs).
• T3.4: Dissemination (experimental results, workshops, publications, demos).

Scientific Methodology

• Evidence-driven baseline: literature and technology scouting to define assumptions and measurable objectives.
• Controlled PoC evaluation: functional and stress testing on model scenarios, including correctness of configurations, coherence of generated documentation, and accuracy of environmental analysis.
• Auditability and safety: explainable outputs and human-in-the-loop control for sensitive actions; policy and constraint adherence.
• Replicability: modular components and clean interfaces to facilitate industrial adoption.

The Team

GAINS is delivered by a multidisciplinary team combining telco/cloud operations, AI engineering, research methodology, project management, and governance.

The project plan also includes additional PhD-level consultant to further strengthen specialized competencies and academic dissemination capacity.

Industrial Partner - WARIAN SRL

WARIAN SRL is an Italian Managed Infrastructure Provider founded in 2011 (HQ: Mercato San Severino, SA and Montoro, AV). The company specializes in fiber-based connectivity, complex network operations, and cloud-integrated services, with strategic interconnection points in Italy, Europe and Latin America (e.g., NAMEX, MIX, AMS-IX among others). WARIAN operates as an innovative SME, maintains an internal R&D team, and has been part of the DHH Group since 2022.

• Operational footprint: a European-wide infrastructure under a single ASN (AS56911), supporting realistic large-scale scenarios.
• Automation maturity: internally developed automation platform integrating order/ticket tracking and network operations.
• Data capabilities: AI-optimized Big Data address database (80+ million addresses) used in production.
• Dissemination track record: industry events and open innovation initiatives (e.g., EnableX), plus academic connections and publications.

Learn more: www.warian.net  |  AS56911 on PeeringDB

Sustainability Focus and DNSH

GAINS treats sustainability as an operational engineering target, not a reporting afterthought. The project aims to reduce avoidable energy waste by minimizing over-provisioning, enabling dynamic resource allocation, and supporting predictive maintenance and proactive fault management. Expected benefits include faster deployment and maintenance cycles, fewer manual errors, improved performance optimization, reduced downtime, and stronger sustainability reporting readiness.

GAINS is aligned with the Do No Significant Harm (DNSH) principle and addresses the six EU environmental objectives through AI-driven optimization and operational best practices:

• Climate change mitigation: lower energy consumption via optimization and intelligent handling of inactive resources.
• Climate change adaptation: improved resilience through predictive analytics and proactive fault management.
• Water and marine resources: no direct impact; indirect reduction through efficiency and use of certified data centers (ISO 30134:2016).
• Circular economy: reuse of standardized configurations and extension of existing hardware lifecycle.
• Pollution prevention: reduced waste and emissions by minimizing operational errors and rework.
• Biodiversity and ecosystems: no direct intervention; indirect reduction of physical footprint by limiting unnecessary infrastructure growth.

Open Science and FAIR Data

GAINS supports Open Science principles and FAIR data management by promoting timely dissemination, transparent documentation, and (where possible) open-access outputs and reusable knowledge artifacts.

Outputs and Access

• Public Abstracts - open dissemination
• Materials - brochure, one-pagers, media assets
• Events - conferences and workshops where GAINS is presented
• Get in touch - request full paper access, collaboration, or speaking invitations