Faysal AzizAdvanced AI & ML with Python
for Materials Engineers From fundamentals to autonomous agents and failure detection in materials imaging - 5 days
Why this course
Predictive technologies and computer vision are revolutionizing materials engineering, enabling unprecedented precision in design, analysis, and quality control. This intensive course builds from Python fundamentals and machine learning principles to advanced agentic AI systems and deep learning architectures specifically engineered for automated failure detection in visual data.
Students will master critical applications including defect identification in material microstructures and circuit-in-film analysis, working with real-world datasets that mirror current industrial challenges. Under the guidance of an instructor with three decades of hands-on industry expertise, participants will develop proficiency in Google Colab environments and establish robust Python workflows on Linux server architectures—skills that directly translate to professional engineering practice.
The curriculum bridges theoretical foundations with practical implementation, preparing engineers to deploy cutting-edge AI solutions in materials science and manufacturing quality assurance.
Learning outcomes
By the end of the course, students will be able to:
- Apply Python basics and libraries (NumPy, Pandas, Matplotlib, Scikit‑Learn) for data exploration and ML workflows.
- Build and evaluate supervised and unsupervised ML models in Colab and on Linux.
- Understand and implement agentic AI systems with dynamic replanning, multi‑step tasks, and failure analysis.
- Design and train deep learning models for image‑based failure detection in materials science contexts.
- Deploy and experiment with pipelines for automatic defect identification from microstructure or cipher MVC images.
Prerequisites
- Basic programming exposure (any language).
- Engineering or materials science background.
- Access to Google Colab and a Linux server; ability to install Python and common libraries (pip, virtualenv).
- Familiarity with basic probability/statistics is helpful.
3 modules
01Python & Machine Learning Foundations4 topics
Python Programming Essentials
- Variables, data types, control flow, functions, modules in Python
- Using Jupyter notebooks in Google Colab and Linux environments
Data Manipulation & Visualization
- NumPy arrays, Pandas DataFrames, data cleaning & transformation
- Plotting with Matplotlib/Seaborn, exploratory data analysis
Classification & Regression Algorithms
- Linear regression, logistic regression, decision trees, support vector machines, k‑nearest neighbors
- Loss functions, regularization, model metrics (accuracy, precision/recall, R²)
Unsupervised Learning & Feature Engineering
- Clustering (K‑means, hierarchical), dimensionality reduction (PCA, t‑SNE)
- Feature scaling, encoding, cross‑validation, hyperparameter tuning
02Agentic AI: Autonomous Systems and Intelligent Agents4 topics
Introduction to Agentic AI
- Definition: fully autonomous agents making decisions without human intervention
- Comparison to rule‑based automation and RPA
Core Components of Agentic Agents
- Memory systems: working, episodic, long‑term using vector databases
- Tool interface layer: APIs, file operations, external services
- Dynamic replanning: detect failures, analyse cause, generate alternative strategies
Failure Modes in Agentic AI
- Novel agentic‑specific failures: inter‑agent communication breakdowns, divergence from task logic
- Classical AI failures: bias, hallucinations, misclassifications
Hands‑on: Building an Agent Pipeline in Python
- Set up agent that monitors a process (e.g. file inputs), triggers ML task, adapts when outputs fall below threshold
- Logging, recovery strategies, continuous learning loops
03Deep Learning for Image‑based Failure Detection5 topics
Deep Learning Fundamentals
- Overview of convolutional neural networks (CNNs), activations, pooling layers, training loops
- Frameworks: TensorFlow/Keras or PyTorch on Colab/server
Image Preprocessing & Data Augmentation
- Techniques to prepare microscopy or cipher MVC images: normalization, resizing, augmentation (rotation, flipping)
Model Architecture Design
- Standard CNN designs (e.g. simple CNN, ResNet variants), transfer learning from pre‑trained models
Training, Validation & Evaluation
- Loss functions for classification/localization, confusion matrices, ROC/AUC metrics
- Techniques to avoid overfitting: regularization, dropout, augmentations, early stopping
Model Explainability & Robustness
- Saliency maps, Grad‑CAM, understanding why models succeed/fail
- Adversarial robustness in imaging and detection contexts (relevant to quality control)
Instructor’s Experience & Real‑World Emphasis
- All content is grounded in industry‑relevant scenarios: inspection of materials, failure prediction, automated quality control.
- Hands‑on labs and examples drawn from real datasets used in aerospace, semiconductor, and materials processing.
- Emphasis on practical pipelines—from data acquisition through deployment—with reproducible code across Colab and Linux.
- No academic fluff: you get tools, solutions, and workflows that matter in today’s engineering environments.
A programme built around your team.
Share your training goals and requirements.