COURSE
RECENT ADVANCES IN
RECENT ADVANCES IN
ML/AI MODEL DEVELOPMENT AND VALIDATION
Mastering Modern ML: From Transparent Models to Trustworthy Validation
RECENT ADVANCES IN
ML/AI MODEL DEVELOPMENT AND VALIDATION
Mastering Modern ML: From Transparent Models to Trustworthy Validation
ABOUT THE COURSE
A hands-on workshop facilitated by MoDeVa focused on recent advancements in machine learning (ML) and artificial intelligence (AI). This workshop is carefully structured to provide in-depth knowledge and practical skills for professionals and enthusiasts interested in the latest techniques for ML model development and validation.
Participants will gain insights into building inherently interpretable machine learning models, using cutting-edge methods such as Gradient Boosted Trees, Neural Trees, GAMI-Net, and Mixture of Experts. The session will also cover advanced methods of explainability both to design inherently interpretable machine learning via Functional ANOVA as well as post-hoc approaches
The workshop will emphasize rigorous validation approaches to ensure the reliability, robustness, and resilience of ML models. Attendees will explore conceptual soundness, model hacking techniques, advanced residual analysis, and techniques for identifying and clustering model weaknesses and failures.
Robustness diagnostics for overcoming overfitting. Resilience diagnostics to anticipate the impacts of distribution drift . Reliability diagnostics to identify decision uncertainty.
An important segment of the workshop will address fairness and debiasing in machine learning. Participants will learn how to measure and diagnose biases, apply effective de-biasing strategies, and design ethical ML solutions through thoughtful feature engineering and optimization.
Lastly, the workshop offers extensive insights into generative language model validation. Attendees will discover the intricacies of embedding models of contrastive and classifier training, automated test generation methods, and the evaluation of semantic similarity and natural language inference. Model weakness identification, robustness and adversarial testing methods will also be thoroughly discussed, preparing participants to address critical challenges in the deployment of generative AI systems.
Learning Objectives
Develop and Validate Interpretable Machine Learning Models
Apply Advanced Validation, Diagnostic, and Monitoring Techniques
Ensure Fairness, Ethical Integrity, and Robustness in Generative AI Systems
Lead Instructor
Location & Dates
Workshop Date:
July 7–8, 2025
Time: 9:00 AM – 5:00 PM (both days)
Venue:
Dubois Center at UNC Charlotte
320 E 9th Street, Charlotte, NC 28202
All attendees must check in at the Dubois Center lobby upon arrival to complete registration and receive name badges. Please plan to arrive at least 30 minutes early to allow time for check-in.
Agenda
- Developing Interpretable Machine Learning
- Introduction to Explainable Machine Learning
- Post-hoc Explainability: SHAP and PDP
- Inherently Interpretable ML via Functional ANOVA Representation
- Building Inherently Interpretable ML
- Gradient Boosting Machine
- Gradient Boosted Linear Tree
- Neural Tree
- Deep ReLU Networks and GAMI-Net
- Mixture of Experts
- Introduction to Explainable Machine Learning
- Validating Machine Learning Model
- Introduction to Model Validation
- Conceptual Soundness Investigation
- Model Hacking: Identification of Hidden Weakness
- Advanced Residual Analysis
- Weakness Detection via Inherently Interpretabale ML
- Failure Clustering
- Resilience Diagnostics: Effects of Distribution Drift and Model Monitoring
- Robustness Diagnostics and Overfitting
- Reliability Diagnostics and Uncertainty
- Fairness and Debiasing
- Introduction to Fair Machine Learning and Measurement
- Fairness Diagnostics
- De-biasing strategy
- Building Less Discriminatory Alternatives
- Objective Function Trade-Offs
- Hyperparameter Optimization
- Feature Engineering
- Validating Generative Language Model
- Introduction to GenAI Model Validation
- Introduction to Embedding Model and Training
- Contrastive Embedding
- Classifier Embedding
- Automated Test Generation: Topic Modeling and Query Types
- Model Evaluation: Sematic Similarity and Natural Language Inference
- Human Calibration: Sampling, Calibration Model and Conformal Prediction
- Identification of Model Weakness: Marginal Analysis and Failure Clustering
- Robustness Test and Evaluation: Adversarial, Out of Distribution and Sensitivity
Pricing Options
This hands-on workshop is priced per participant, with significant discounts available for organizations that are members of the TAIMing AI Center. Membership not only reduces the cost per attendee but also provides year-round access to other trainings, workshops, and research collaboration opportunities. Volume discounts apply for companies sending multiple participants, which stack with membership rates for maximum value. To discuss becoming a member, please contact us.
| Membership Level | 1–4 Participants | 5–9 Participants (10% off) | 10+ Participants (15% off) |
|---|---|---|---|
| Non-Member | $1,500 / person | $1,350 / person | $1,275 / person |
| Silver Member ($25K) | $1,125 / person | $1,012 / person | $956 / person |
| Gold Member ($50K) | $900 / person | $810 / person | $765 / person |
| Platinum Member ($150K) | $750 / person | $675 / person | $637 / person |
Pre-reading Materials
Pre-reading materials will be provided at a later date.
Security/Event Check In Requirements
All attendees—University and Non-University—must report to the Dubois Center lobby upon arrival to the property to complete visitor registration and receive name badges before proceeding to the classroom. A representative of the Center for TAIMingAI will be set up in the lobby to assist attendees with the check-in process. Representatives of the TAIMingAI Center are expected to arrive 30 minutes prior to the start of the event to begin assisting attendees with check-in. However, if a representative of the Center for TAIMingAI is not present upon your arrival, you must wait in the lobby. Faculty of UNC Charlotte must wear their University-issued nametag or other official identification.
*Please note that attendees who do not complete the check-in process will be denied classroom access, and may be asked to leave the property–as outlined by the Dubois Center’s Security Procedures and Event Guidelines—no exceptions.
Dubois Center Security Procedures
FAQs