CHARAL 2025

Challenges in Holistic AI for Real-world Applications and Learning Systems

"From Algorithms and Models to Real-world AI Integration"

Workshop at MICAI 2025

Date: November 4th, 2025

Location: Guanajuato, Mexico. Centro de Investigación en Matemáticas (CIMAT)

Format: Hybrid (In-person & Virtual)

Join our email list for updates!

About CHARAL

As AI technologies mature, the focus is shifting from model development to their effective deployment in real-world systems. In domains such as robotics, manufacturing, healthcare, and intelligent infrastructure, integrating AI into production environments introduces unique challenges related to system design, latency, reliability, monitoring, and trust.

The 1st CHARAL workshop at MICAI 2025 will provide a forum for researchers and practitioners to explore holistic approaches to engineering and operating AI systems that are robust, explainable, and adaptable in complex, real-world conditions.

Systems Engineering for AI Deployment

Monitoring & Observability

Trust & Safety

Data-Centric AI in Production

Specialized Deployment Challenges (Foundational models, AI on Edge, Robotics)

Workshop Program

Date: November 4th, 2025 | Time Zone: CST (UTC-6)

Time Session Title Presenter
9:00 - 9:20 Introduction Workshop introduction and its goals Chairs
9:20 - 10:10 Keynote Talk Uncertainty in Vision Language Models and Computer Vision Dr. Matias Valdenegro
10:10 - 11:00 Keynote Talk Deep Generative Modeling for Multimodal Human Trajectory Prediction Dr. Jean-Bernard Hayet
11:00 - 11:20 Spotlight Talk A Metamodelling Framework for Accelerated Energy Market Optimization using Active Learning TBA
11:20 - 11:30 COFFEE BREAK
11:30 - 11:40 Lightning Talk On the Interpretation of Clip Limit in Contrast-Limited Adaptive Histogram Equalization TBA
11:40 - 12:30 Keynote Talk Challenges in Driving Perception: How to Solve It with Open-Source Solutions (Voxel 51) Paula Ramos, PhD
12:30 - 12:50 Spotlight Talk Track Almost Everything Underwater: Evaluating Visual Registration in Marine Robotics TBA
12:50 - 14:00 LUNCH
14:00 - 14:10 Lightning Talk Adaptation of Pre-trained Neural Network Models for Inspection via Image Analysis TBA
14:10 - 15:00 Keynote Talk Early Detection of Diabetic Retinopathy Dr. Ulises Moya
15:00 - 15:50 Keynote Talk A Clinical Framework for Real-World Medical AI Diego Cesar Lerma Torres, M.D.
15:50 - 16:10 Spotlight Talk Drupelet: End-to-End Development of Mixed Binary/Ternary Neural Networks for Ultra Resource-Constrained Microcontrollers TBA
16:10 - 16:20 COFFEE BREAK
16:20 - 16:30 Lightning Talk Agent Cards: A Documentation Standard for Operational AI Agents TBA
16:30 - 17:20 Keynote Talk Efficient Training of Large Language Models (LLMs) Omar U. Florez, PhD
17:20 - 18:10 Panel Discussion AI in the Wild, From Scholar/Lab Models to Living Systems Chairs & Young AI Leaders Mexico
18:10 - 18:20 Conclusion Closing remarks Chairs

Keynote Speakers

Omar U. Florez, PhD
Omar U. Florez, PhD
"Efficient Training of Large Language Models (LLMs)"
National Center of Artificial Intelligence (CENIA), Chile
Paula Ramos, PhD
Paula Ramos, PhD
"Data-Centric AI and Open-Source Tooling for Safer, Trustworthy AI Integration"
Voxel 51
Dr. Ulises Moya Sanchez
Dr. Ulises Moya Sanchez
"Implementing End-to-End AI Solutions in Jalisco’s Government"
Director IA del Gobierno de Jalisco
Dr. Matias Valdenegro
Dr. Matias Valdenegro
"Uncertainty in Vision Larnguage Models and Computer Vision"
University of Groningen
Dr. Jean-Bernard Hayet
Dr. Jean-Bernard Hayet
"Deep Generative Modeling for Multimodal Human Trajectory Prediction"
Centro de Investigación en Matemáticas, A.C.
Diego Cesar Lerma Torres, M.D.
Diego Cesar Lerma Torres, M.D.
"A Clinical Framework for Real-World Medical AI"
Neuroglia Solutions

Organizing Team

Chair Photo

Ing. Arturo Gomez Chavez

Workshop Chair
Constructor University Bremen
Co-Chair Photo

Dr. Emmanuel Ovalle-Magallanes

Co-Chair
Universidad de La Salle Bajio
PC Member Photo

Dr. Jose Martinez Carranza

Co-Chair
Instituto Nacional de Astrofisica Optica y Electronica (INAOE)

Call for Papers

We invite researchers and practitioners to submit papers addressing the challenges of deploying AI systems in real-world environments. We welcome both theoretical contributions and practical case studies. The technical paper should be written in English in LNCS Springer style, not exceeding 12 pages. The submissions must not contain author's names or affiliations.

The accepted papers will be published in the Springer LNAI Proceedings, and the author fee amount for presentation at MICAI conference and publication is $5,000 MXN (approximately $300 USD). Submission should be made through the CMT system at:

Submit Your Paper

Important Dates

All deadlines are at 11:59 PM PST Time

Submissions will be reviewed on a rolling basis to allow for authors to plan their registration and participation ahead.

*Early bird submission allows for the author to register to MICAI for early-bird pricing

Any questions of inquiries, please send us an email to support@charal-workshop-ai.com

*EARLY BIRD REGULAR
Submission September 12th, 2025 October 10th, 2025
Acceptance Notification September 18th, 2025 Ocotber 16th, 2025
Camera-Ready Submission October 22th, 2025
Workshop Date November 4th, 2025

Systems Engineering for AI Deployment

Architectural patterns, system constraints management, fail-safe design, and graceful degradation in AI pipelines.

Monitoring, Observability & Operational Feedback

Observability instrumentation, logging and tracing, debugging techniques, and performance feedback loops.

Trust, Safety & Assurance in Deployed Models

Uncertainty estimation, explainability methods, red-teaming, adversarial robustness, and evaluation metrics.

Data-Centric AI in Production

Real-world data pipelines, feedback-driven collection, active learning, and weak supervision techniques.

Foundation Models in Production

Adapting large models for latency-sensitive environments, robotics, and manufacturing systems.

AI on Edge & Embedded Systems

Quantization, pruning, distillation, streaming inference, and federated updates for resource-constrained environments.