Predictive Modelling and Risk Assessment with Machine Learning and Deep Learning
A three-day workshop for civil engineers, from interpretable models to LSTM forecasting
A three-day live online workshop with Dr. Kaushik Jas: Python and the mathematics of AI, then liquefaction potential with gradient boosting and SHAP, then LSTM time-series forecasting. Hands-on in Google Colab, open to undergraduates with no prior coding.
No coding experience needed: day one starts from Python basics.
- 3
- Live sessions
- 90–120 min
- Per session
- 4.5–6 hrs
- Live instruction
- Google Colab
- Code in browser
- 11–13 Dec
- 5:30 p.m. IST
- Live
- Online format
- 2 weeks
- Recording access
- Civil
- Engineering focus
You won’t install Anaconda, MATLAB or any other desktop software. Every hands-on session runs in Google Colab in your browser, and you keep the notebooks.
One engineering workflow, start to finish
Follow one practical process from an engineering question to a prediction you can explain.
- 01Frame the engineering riskLiquefaction, ground motion, soil response.
- 02Understand the data and the mathsLinear algebra, probability and calculus, in brief.
- 03Train ML and DL modelsFrom regression to boosting to LSTM.
- 04Evaluate performanceTest every model against unseen data.
- 05Explain the predictionsFeature importance and SHAP, not a black box.
- 06Assess the riskTurn a prediction into an engineering judgement.
Machine learning is now routine in civil engineering research, but a model nobody can explain is hard to trust with a foundation, a slope or a city’s seismic risk. This workshop builds predictive models for engineering risk, and shows how to check and justify what they predict.
Day one gives everyone the same footing, from what AI, ML and deep learning are to the mathematics behind them and Python in Google Colab. Day two builds and explains gradient-boosting models for liquefaction potential, Dr. Kaushik’s Ph.D. research area. Day three moves to deep learning, forecasting time histories with LSTM networks, and surveys applications from structural health monitoring to water resources and remote sensing.
Who it’s for
- 1st- to 3rd-year B.Tech./B.E. Civil Engineering students meeting Python and ML for the first time
- Final-year, M.Tech./M.E. and Ph.D. students starting ML research
- Faculty members who want to teach or supervise ML work in civil engineering
- Geotechnical, structural and earthquake engineers working with data
- Researchers who need models they can explain to reviewers and clients
No prior Python or Machine Learning is required: day one covers the basics. Comfort with school-level mathematics helps, and the mathematical foundations are introduced in the first session.
What you’ll be able to do
The foundations
- Understand the mathematical foundations and working principles of AI, machine learning and deep learning algorithms.
- Write and run Python in Google Colab, from the basics to a working ML pipeline.
Models for real problems
- Develop and implement machine learning and deep learning models for civil engineering problems.
- Assess liquefaction potential as a classification problem with GBM and XGBoost, using code you can reuse on your own data.
Predictions you can defend
- Interpret and explain model predictions with feature importance and SHAP.
- Evaluate the predictive performance of AI-based models for engineering risk assessment, and forecast time histories with an LSTM network.
3 live sessions
Each day pairs the ideas and mathematics with a hands-on session in Google Colab.
Foundations
AI, its mathematics and Python
01Day 1Introduction to Interpretable AI, Its Mathematical Foundations and PythonLearn what AI, ML and DL are, the maths behind them, and write your first Python in Colab.
AI, ML and DL · Supervised, unsupervised and reinforcement learning · Interpretable AI · Linear algebra · Probability and statistics · Calculus · Python fundamentals · Google Colab
AI and ML in everyday life and civil engineering, including computer vision
Risk assessment
Boosting and SHAP for liquefaction
02Day 2Interpretable Machine Learning: Liquefaction Potential with Gradient Boosting and SHAPBuild a liquefaction classifier, test it, and explain what drives each prediction.
Linear and logistic regression · Decision trees · GBM · XGBoost · LightGBM · Explainable Boosting Machines · Training, testing and evaluation · Feature importance · SHAP
Liquefaction potential assessment of soils, as a classification problem
Deep learning
LSTM forecasting and applications
03Day 3Deep Learning for Civil Engineering: Time-Series Forecasting with LSTMTrain an LSTM network to forecast a time history, with code you can carry into your own work.
Recurrent neural networks · LSTM · Time-history prediction · Interpretable ML and DL in practice · Research case study
Soil response time histories, structural health monitoring, ground motion, construction management, water resources and remote sensing
Hands-on sessions run on demonstration datasets in Google Colab. Dr. Kaushik shows the liquefaction database from his research and points you to public sources of laboratory data; unpublished laboratory data isn’t shared.
Ideas, then code, then questions
90 to 120 minutes each day: the ideas first, then code you run yourself, then your questions.
Do the work with us
- 01Live lectures that explain the mathematics behind each algorithm
- 02Hands-on coding in Google Colab in every session
- 03Real civil-engineering problems: liquefaction and soil time histories
- 04Interpretable models and SHAP explanations rather than black boxes
- 05Your questions answered live in every session
Leave with a working toolkit
- The Google Colab notebooks from all three sessions
- Pointers to public liquefaction and laboratory data sources
- Reusable code for classification and time-series forecasting
- A suggested reading list, from Bishop to Molnar
- Live question time with the instructor every day
- Session recordings, available for two weeks after the workshop
- Certificate of participation
Nothing to install
Everything else is ready
- A Google account and a web browser
- Google Colab as the teaching environment
- Python 3 with NumPy, pandas and Matplotlib
- scikit-learn, XGBoost and LightGBM
- Explainable Boosting Machines and SHAP
- A deep-learning library for the LSTM session, ready in Colab
Built for engineers, not data scientists
Risk you can justify
Every prediction comes with the evidence behind it, from feature importance to SHAP.
Civil engineering problems
Liquefaction and soil time histories, not generic toy datasets.
From ML to deep learning
Regression and boosting through to LSTM networks, in three days.
Taught by the researcher
Dr. Kaushik teaches the methods from his own published work on liquefaction and cyclic soil behaviour.
Open to beginners
Day one starts from Python basics and the mathematics, so undergraduates can follow along.
Reusable code
Take the notebooks into your own classification and forecasting problems.
Meet your instructorDr. Kaushik Jas
Postdoctoral Research Associate, School of Civil Engineering, Sun Yat-sen University, China
MS–Ph.D. in Geotechnical Engineering, IIT Madras · Prime Minister’s Research Fellow
Dr. Kaushik Jas is a Postdoctoral Research Associate at the School of Civil Engineering, Sun Yat-sen University, Guangzhou, China, where he works on intelligent modelling of underground infrastructure. His research applies interpretable machine learning and deep learning to earthquake geotechnical engineering: liquefaction potential assessment, the cyclic response of liquefiable sands, and geohazard risk.
He completed his MS–Ph.D. in Geotechnical Engineering at IIT Madras as a Prime Minister’s Research Fellow, supervised by Prof. G. R. Dodagoudar, with a thesis on assessing the liquefaction potential of saturated cohesionless soils using machine learning. His explainable XGBoost-SHAP model for liquefaction potential was listed among the top-cited articles of 2025 in Soil Dynamics and Earthquake Engineering. He was a visiting Ph.D. researcher at Hokkaido University, Japan, and has collaborated with Montana State University and JPMorgan Chase & Co.
Full profile, publications & talksBefore you register
When is the workshop?
Friday 11, Saturday 12 and Sunday 13 December 2026, from 5:30 p.m. IST each day, for 90 to 120 minutes. All three sessions are live online.
Do I need to know Python or Machine Learning?
No. Day one covers Python basics in Google Colab and introduces ML from the ground up, so 1st- to 3rd-year undergraduates can join. Days two and three build on it.
How much does the workshop cost?
In India, ₹1,050 for students and Ph.D. scholars and ₹2,100 for faculty and working professionals, all taxes included. Elsewhere the fee is shown in your local currency when you register. The fee covers all three live sessions, the notebooks, two weeks of recordings and the certificate.
Can I get a refund?
The fee is non-refundable. If you miss a session, its recording stays available for two weeks after the workshop.
Will I get the liquefaction database?
Dr. Kaushik shows the liquefaction database from his research during the session, and the hands-on work runs on similar demonstration datasets in Colab. You keep the code to use on your own data. Unpublished laboratory data can’t be shared, but he points you to public sources.
What if I miss a session?
Session recordings stay available for two weeks after the workshop.
Will I get a certificate?
Yes. Participants receive a certificate of participation.
Do I need a powerful laptop?
No. Everything runs in Google Colab in your browser, so any modest laptop works.
Save your seat in the next cohort.
Register in about a minute. Your seat is confirmed as soon as your payment is.
