This year’s competition has drawn an incredible array of ideas from universities, research institutes, startups, and established companies worldwide. The selected teams have proposed innovative solutions in six key focus areas: AI in network technology, AI tools for fiber rollout, AI for sustainable power consumption, AI capabilities for sales and service, AI use cases for network APIs and AI for security.
What’s Next?
Over the coming months, these teams will have the unique opportunity to refine their AI solutions in collaboration with industry experts from Deutsche Telekom and T-Mobile US. Their journey will culminate in a live exhibition and stage presentation at Deutsche Telekom’s headquarters in Bonn on June 2 and 3, 2025.
A distinguished panel of judges will evaluate the solutions, awarding cash prizes totaling up to €300,000. More importantly, participants will have the opportunity to explore business collaborations with Deutsche Telekom and T-Mobile US, potentially scaling their innovations across Germany, Europe, and the U.S.
Join Us in Shaping the Future of Telecom
T Challenge 2025 is more than just a competition—it’s a global platform for innovation, collaboration, and the future of telecom. We invite you to follow the journey of these top 12 teams as they develop AI solutions that could redefine the industry.
Stay tuned for updates as we approach the grand finale in June!
T Challenge Nominees 2025
NeuralAgent
Germany
AgentQoS: Autonomous Service Assurance with Multi-Agent Reinforcement Learning
AgentQoS leverages decentralized AI agents and reinforcement learning to dynamically optimize network performance. By deploying thousands of intelligent agents across routers, base stations, and edge devices, it enables real-time, self-adapting QoS management for next-gen technologies like 5G, AR/VR, and satellite direct-to-cell, ensuring efficient resource utilization and superior service quality.
Nanites AI
USA
AI agents that autonomously troubleshoot and fix networks
AI-driven agents detect, diagnose, and resolve network issues in minutes, reducing MTTR by up to 90%. Acting like human engineers, they analyze alerts, access devices, troubleshoot problems, and take corrective action in real time. This enables faster, more reliable, and cost-efficient network operations for service providers, data centers, and
enterprises.
Rutgers University
USA
RescueNet: AI-Driven Radio Access Network for Rapid Disaster Response
RescueNet is a portable, AI-driven 5G network designed for rapid deployment in disaster zones. Using Open RAN
architecture and satellite backhaul, it ensures reliable, high-speed connectivity for first responders and affected
communities. Its AI engine optimizes spectrum, manages traffic, and enables real-time edge computing, providing
fast, efficient, and resilient communication in critical situations.
Themis AI
USA
Capsa Enhances AI Models, Catching Errors and Improving Performance
In seconds, Capsa automatically enhances any AI model, producing an updated version that reports its own uncertainty. This enables efficient correction of dataset errors and flagging of uncertain outputs in real time. A more robust alternative to traditional methods of uncertainty estimation, Capsa can improve network planning, automation, and AI-driven decision-making.
Shield IOT
Israel
AI Network Security Platform For Securing Mass-Scale 5G & Cellular IoT Networks
Shield-IoT is an agent-less AI platform designed to secure mass-scale 5G and cellular IoT networks like connected cars, smart grids, and smart cities. It provides real-time threat monitoring, asset visibility, and compliance management without requiring any network changes. Using patented Coreset-AI technology, it offers scalable, privacy-ready protection for billions of SIM-connected devices, detecting cyber threats like malware and device takeovers with industry-first anomaly detection.
Exeon Analytics
Switzerland
AI-powered network security
Exeon offers AI-driven Network Detection & Response (NDR) technology to protect companies from cyberattacks. By analyzing network metadata, our platform detects zero-day threats, multi-stage attacks, and network anomalies
without additional sensors or agents. Integrated into SASE products, it enables early detection of breaches and data
exfiltration, improving the cyber resilience of companies. Exeon’s solution provides advanced behavior analysis, using AI and machine learning to identify known and unknown threats in distributed, hybrid environments.
Resemble AI
USA
Deepfake Detection
The platform leverages AI, audio forensics, and advanced video analytics in a multimodal approach to detect deepfakes in real time across calls, voicemails, virtual assistants, and video streams. It combines voice biometrics with machine learning to accurately identify AI audio manipulations and visual anomalies in facial expressions and audio-video sync. Offering high accuracy with low false positives, it seamlessly integrates into existing authentication systems and monitoring frameworks, helping telecom providers protect against voice phishing, video-based manipulations and fraud.
Hiverge
United Kingdom
Optimizing AI infrastructure energy usage with AI
Hiverge pioneers program synthesis – a technology combining Large Language Models with symbolic systems – to automatically discover algorithms that significantly improve AI operations efficiency. This creates interpretable, verifiable algorithms that automatically adapt to changing environments, overcoming limitations of rule-based heuristics and black-box ML. The technology drives significant energy and cost savings for telecom operators running AI jobs on edge data centers.
Opanga Networks
USA
AI/ML for Dynamic RAN energization based on real-time Customer Experience measurements
Opanga’s AI/ML platform dynamically adjusts RAN energy consumption by analyzing real-time customer experience
(CX) across wireless network cells. The system correlates CX measurements to optimize energy savings by deenergizing
capacity layers when not needed (e.g., during low-demand periods) and proactively re-energizing cells as demand changes. Unlike traditional RAN energy savings features that rely on fixed schedules, Opanga’s solution uses real-time insights into customer demand, achieving up to 40% energy savings with no negative impact on CX.
Space Data AI
USA
FiberDocAI
FiberDocAI is a multi-agent AI platform designed to automate and optimize documentation for fiber rollout projects.
Each AI agent is specialized and trained to handle critical aspects, including design, permitting, construction, and
compliance. By streamlining these processes, FiberDocAI reduces permit approval time by 80%, cuts errors by 90%,
and digitizes 95% of paperwork, ensuring seamless project execution.
Hiya
USA
Hiya AI Voice Detection: Protecting the Contact Center from AIpowered fraud
Hiya AI Voice Detection is an AI-powered solution that accurately distinguishes real voices from AI-generated ones in real-time, with 99.2% accuracy. Designed for contact centers, it integrates via API and can be deployed in cloud,
private cloud, or on-premises. It detects deepfake voice scams, enhancing security and improving efficiency in
handling sensitive customer data, with real-time detection during phone calls. The solution is scalable and optimized for fraud prevention.
Qualimero Technologies
Germany
AI-Sales Workers
Qualimero’s AI sales workers qualify leads, advise on products, schedule appointments, and create offers in multiple
languages, 24/7. Integrated with your CRM, they engage prospects via WhatsApp, phone, and video, boosting
conversions by 3x and cutting acquisition costs by 30%—automating sales while maximizing impact.
