A joint initiative by India’s Department of Telecommunications and the International Telecommunication Union produced 16 AI prototypes aimed at network analysis, security, troubleshooting and resource optimisation, with a Bengaluru engineering team taking the top prize.
India’s Department of Telecommunications (DoT) and the International Telecommunication Union (ITU) have showcased agentic artificial intelligence prototypes designed to address practical challenges in telecom networks. The projects were developed under the AI for Good Build-a-thon 5.0, held as part of India Mobile Congress (IMC) 2026 at Yashobhoomi in New Delhi.
The initiative attracted more than 25 registered teams and resulted in 16 prototype submissions. Team Netaxis from RV College of Engineering (RVCE), Bengaluru, won first prize for its proposal on context-aware orchestration of quality of service and adaptive security in resource-constrained 5G networks.
DoT–ITU Build-a-thon focuses on AI agents for telecom
The Build-a-thon was designed to help students, researchers and professionals develop AI agents for telecommunications use cases, including network troubleshooting, resource optimisation and policy-aware analysis. According to the DoT, the programme began in early July, with participants receiving guidance through five online mentoring sessions conducted by DoT and ITU experts.
The mentoring covered telecom-related problem statements, open-source data sources, the ITU-T Y.3172 machine-learning pipeline and the ITU AI Readiness 2.0 report. Participants also worked with the ITU AI for Good Sandbox to develop solutions grounded in relevant national policies, technical standards and regulatory requirements.
The submissions included knowledge-base agents for specialised telecom problems, policy and legal adviser agents, and coordinator agents designed to manage interactions between multiple AI agents. This approach is intended to move beyond a single chatbot responding to prompts towards systems that can analyse information, use defined tools and coordinate tasks within a controlled workflow.
Bengaluru team wins first prize for 5G network orchestration
Team Netaxis from RVCE Bengaluru secured first place for its project, titled “Context Aware Agentic Orchestration of QoS and Adaptive Security in Resource Constrained 5G”. The concept combines network service management with adaptive security, addressing the challenge of maintaining service quality when network resources are limited.
Quality of service (QoS) management is important in 5G networks because different applications have different requirements for latency, throughput and reliability. A network handling enterprise applications, connected devices and other data-intensive services must allocate resources according to changing traffic conditions and service priorities.
An agentic system could potentially help analyse network conditions, recommend resource allocation and coordinate security responses. However, the project’s award recognises a prototype proposal; it does not establish that the system has been deployed by a telecom operator or demonstrated at commercial scale.
Team SNR received second prize for an approach combining knowledge-grounded retrieval, deterministic constraint validation and multi-agent arbitration. Team Agentnet won third prize for “TeleAgent”, a collaborative multi-agent framework using a fine-tuned small language model for automated troubleshooting of private 5G networks.
Other prototypes target network security and automation
The DoT highlighted a solution for multi-agent network analysis with policy mapping. It combines a country-specific 5G and telecom knowledge base with AI-based analysis of network-error scenarios, mapping to policies and standards, identification of gaps, human supervision and auditable evidence.
Such an approach could help telecom engineering and compliance teams connect technical troubleshooting with the rules and standards applicable to a particular network environment. Keeping a human supervisor involved and maintaining an auditable record are important considerations when AI-generated recommendations may affect network configuration or security.
The judges’ prize went to Team Nethealers for an autonomous self-healing 5G radio access network agent. Its proposal uses foundation-model-based diagnosis and digital-twin pre-validation, an approach intended to assess proposed corrective actions in a simulated environment before they are considered for a live network.
The special women’s prize was shared by Team Orbitflow from SNU Chennai for predictive traffic-aware beam hopping for low-Earth-orbit non-terrestrial networks, and Team Handshake Protocol for law-aware AI agents for 5G emergency response.
Why agentic AI matters for telecom operators
Telecom networks generate large volumes of performance, fault and security data. Engineers must interpret these signals, identify the cause of service degradation and decide which corrective action is appropriate. AI agents that can retrieve relevant technical information, analyse scenarios and coordinate defined tasks could help reduce some of this manual work.
Potential applications include identifying recurring faults, supporting private 5G network operations, recommending traffic-management changes and helping engineers evaluate network policies. Multi-agent systems may be particularly useful when a task requires separate forms of analysis, such as technical diagnosis, regulatory checks and validation of a proposed response.
The practical value will depend on the accuracy of the underlying data, the reliability of each agent, integration with network-management systems and the safeguards governing automated actions. Telecom networks also have strict availability and security requirements, making testing, access controls, human oversight and rollback mechanisms essential before such systems are used in production.
Policy-aware AI is part of the initiative
The programme placed emphasis on grounding solutions in Indian policies, standards and regulations rather than developing generic AI demonstrations. This is relevant for telecommunications, where network operations can involve cybersecurity requirements, privacy obligations, service commitments and infrastructure constraints.
The prototypes provide examples of how agentic AI might support telecom operations, but the next step is to establish their performance under realistic network conditions. Independent testing, integration with operator systems and evidence of reliability would be needed to determine whether the concepts can deliver measurable operational benefits.
The DoT–ITU Build-a-thon therefore offers a view of emerging Indian work at the intersection of AI and telecom engineering. Its immediate outcome is a set of prototypes and use cases, with the longer-term opportunity lying in whether these ideas can be validated and translated into dependable tools for real-world networks.
