We propose a presentation on our TME-AIX project, showcasing practical AI implementations in telecommunications using PyTorch on consumer-grade hardware. Our work demonstrates how telcos can effectively implement classification, regression, clustering, and anomaly detection models without requiring specialized infrastructure. Key Points:
(1)Implementation of "mixture of experts" approaches combining traditional ML with transformers
(2) Real-world telco use cases: fraud detection, service assurance, traffic segmentation, and security anomaly detection
(3) Data engineering using OpenTelemetry for telco-specific datasets
(4) Complete open-source implementation with notebooks and models available on GitHub and Hugging Face

Our presentation will provide practical insights for implementing cost-effective AI solutions in telecommunications environments with PyTorch, demonstrating that effective AI adoption doesn't necessarily require massive computational resources.