The short course is open to all and is tailored to introducing attendees to Machine
Learning and Artificial Intelligence as applied to the field of Geotechnical Engineering. It aims to provide the practical skills needed to apply artificial intelligence (AI) in the field. The curriculum integrates industry-standard tools such as Python, PyTorch, SciKit-Learn, SHAP, Pyro, and HuggingFace, ensuring participants can master both the fundamentals and advanced techniques of machine learning and deep learning for practical geotechnical applications. Each module has hands-on coding exercises directly tied to real-world geotechnical challenges for participants to execute at home.

Dr. Kumar will offer some (but not all) of the following modules during the training. An updated outline is forthcoming.

• Module 1: Introduction to Machine Learning for Geotechnical Engineering. Establish ML basics and relevance to geotechnical problems. Supervised vs. unsupervised learning, regression, classification, geotechnical use cases. Hands On: Exploratory data analysis (EDA) on borehole data; train a logistic regression model using SciKit-Learn for soil type prediction.
• Module 2: Data Preprocessing, Feature Engineering, & Basic Models. Clean data and implement foundational ML models. Handling missing data, normalization, decision trees, linear/logistic regression. Hands-On: Clean landslide data; build a decision tree classifier using SciKit-Learn for susceptibility mapping.
• Module 3: Explainable AI (XAI) with Random Forests for Landslide Risk. Predict landslides with interpretable models. Random Forest theory, SHAP values, feature importance.
  Hands-On: Train a Random Forest model and interpret results using SHAP.
• Module 4: Clustering for Soil Site Profiling & Layering. Automate soil stratification from CPT/SPT data. Topics include k-means, DBSCAN, PCA for dimensionality reduction. Hands-On: Cluster soil layers using SciKit-Learn and visualize 3D stratigraphy.
• Module 5: Deep Neural Networks (DNNs) & Activation Functions. Design DNNs for geotechnical applications. Neural network architecture, activation functions (ReLU, sigmoid), backpropagation. Hands-On: Build a PyTorch DNN to predict soil compaction from lab data.
• Module 6: LLMs & RAG for Geotechnical Knowledge Management. Build AI-augmented tools for documentation and knowledge management. Retrieval-Augmented Generation (RAG), vector databases, fine-tuning. Hands-On: Create a Q&A chatbot for geotechnical codes using HuggingFace Transformers.
• Module 7: Future Trends & Challenges in AI for Geotechnics. Prepare for emerging AI applications in geotechnical engineering. Ethics, data privacy, IoT integration, digital twins, AI-assisted designs. Hands-On: Case study analysis focusing on AI failures in geotech and developing mitigation strategies.

Upon completing the course, participants will be able to:

(1) Integrate AI techniques into geotechnical engineering practices

(2) Have a basic understanding on how to develop and deploy machine learning models to predict soil behavior, assess risk, and optimize engineering operations.

(3) Interpret AI model outputs using explainable AI techniques and improving decision-making. 

(4) Stay ahead of emerging trends in AI applications within the geotechnical field.