Centram Centre for Applied Artificial Intelligence

AI & MACHINE LEARNING FOR ICT PROFESSIONALS COURSE

Course title

AI & Machine Learning for ICT Professionals

Target audience:

ICT professionals (networking, systems, cloud, security, operations), technical managers, and developers working in telecom, enterprise IT, cloud, or service providers.
Prerequisites
Basic programming (Python recommended),

Foundations in statistics and linear algebra (basic familiarity)
Familiarity with Linux, cloud basics (AWS/GCP/Azure), networking concepts
Recommended pre-course reading: Python basics, Numpy/Pandas tutorials.

Course objectives

By course end, participants will be able to:

  1. Explain core ML/AI concepts and how they apply to ICT problems.
  2. Build, evaluate, and deploy ML models for ICT use cases (anomaly detection, traffic classification, capacity forecasting, predictive maintenance).
  3. Integrate ML models into cloud, container, and networked environments.
  4. Implement MLOps best practices for monitoring, retraining, and governance.
  5. Understand ethical, privacy, and security implications of AI in ICT.

Course duration and modes of delivery

Intensive 4 weeks physical training OR 2 weeks of physical and 3 weeks of online training

Course Content

Introduction to AI/ML for ICT

  1. What is AI/ML? Supervised/unsupervised/reinforcement learning
  2. ICT-specific applications and success stories
  3. End-to-end ML workflow
  4. Python for Data Science (refresher)
  5. NumPy, Pandas, visualization (Matplotlib/Seaborn)
  6. Data wrangling with logs, NetFlow, SNMP, and telemetry.

Data collection & feature engineering in ICT

  1. Network/telemetry data sources, time series, categorical encodings
  2. Aggregation, windowing, labeling, imbalance handling

Supervised learning for ICT

  1. Linear/logistic regression, decision trees, ensemble methods (RandomForest, XGBoost)
  2.  Use cases: traffic classification, QoS prediction, security event classification

Unsupervised learning & anomaly detection

  1. Clustering (k-means, DBSCAN), PCA, isolation forest, autoencoders
  2.  Use cases: intrusion detection, fault detection, rogue device identification

Time series forecasting

  1. ARIMA basics, exponential smoothing, state-space models, LSTM/RNN, Prophet
  2. Use cases: capacity planning, demand forecasting, SLA predictions

 Deep learning basics & specific ICT use cases

  1. Neural network fundamentals, CNNs and RNNs when relevant
  2. Packet/flow analysis, traffic fingerprinting, voice/VoIP analytics

 MLOps & model deployment in ICT environments

  1. Model versioning, CI/CD for ML, containerization (Docker), model serving (REST, gRPC)
  2. Deployment on cloud, edge devices, network function virtualization (NFV)

Monitoring, maintenance & model lifecycle

 Drift detection, re-training strategies, model performance monitoring

  1. Logging, alerting, A/B testing and rollback

Security, privacy & ethics

  1. Adversarial ML basics, privacy-preserving ML (differential privacy, federated learning)
  2. Regulatory concerns and responsible AI for networked systems

Edge & cloud-native ML for ICT

  1. Accelerators (GPUs/TPUs), Kubernetes, serverless, edge deployment patterns
  2. Resource-constrained model optimization (quantization, pruning)

Project & case studies

  1. Real-world case studies (telecom churn/forecasting, IDS, anomaly detection for routers, predictive maintenance)
  2.  Capstone project presentations




Hands-on labs & practical exercises

  1. Parsing and feature extraction from NetFlow/IPFIX/PCAP and syslog
  2. Traffic classification using ML on flow-level features
  3. Anomaly detection on time-series telemetry (router CPU, interface counters)
  4. Forecasting bandwidth demand and SLA breach probability
  5. Building a model serving pipeline with Docker, FastAPI/Flask and Kubernetes
  6. Demonstrating drift detection and automated retraining
  7. Simple adversarial example demonstration and mitigation

Capstone project ideas

  1. Real-time anomaly detection pipeline for network telemetry (NetFlow/Prometheus)
  2. Predictive maintenance for telecom equipment using SNMP and alarm history
  3. Traffic classification and QoS-aware routing suggestion system
  4. Capacity forecasting and automated scaling rules for virtualized network functions.
  5. Security event classification and prioritization for SIEM integration
  6. Federated learning across edge routers for distributed intrusion detection

Tools, frameworks & Datasets

  1. Languages/IDEs: Python, JupyterLab
  2. Libraries: scikit-learn, pandas, NumPy, matplotlib, seaborn, TensorFlow or PyTorch, XGBoost, Prophet
  3. MLOps/Deployment: Docker, Kubernetes, Airflow, MLflow, Seldon/TF Serving, FastAPI
  4. Monitoring: Prometheus, Grafana, ELK stack (Elasticsearch/Logstash/Kibana)
  5. Cloud: AWS/GCP/Azure services (Sagemaker, Cloud ML, Vertex AI), serverless
  6. Datasets: CAIDA, MAWI, CIC-IDS, UNSW-NB15, NetFlow/PCAP sample datasets, internal telemetry

Assessment & evaluation

  1. Labs and quizzes: 30–40%
  2. Midterm assignment/case study: 20–30%
  3. Capstone project (design, implementation, presentation, report): 30–40%

Learning outcomes & competencies

  1. Data pipeline skills: ingest, clean, label, and transform ICT telemetry
  2. Modeling skills: select and tune models appropriate to ICT tasks
  3. Deployment skills: containerize, serve, and monitor models in production
  4. Governance: implement logging, explainability, privacy-preserving methods
  5. Soft skills: translate ICT problems into ML tasks, present findings to managers.

Suggested readings & online resources

  1. Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow — Aurélien Géron
  2. Practical MLOps — Noah Gift, Alfredo Deza
  3. Relevant vendor documentation (AWS/GCP/Azure ML guides)
  4. Research papers and industry blogs (CAIDA, Cisco, Google SRE/ML posts)