Wednesday, April 15, 2026. MathWorks brings AI in industry, hands-on low-code tools, safety-critical systems, and a first look at agentic AI with MATLAB to Kiewit Hall. Day 3 of Husker AI Days.

Note: The following is a real-time draft. Check back for a refined version.

At a Glance

Headline

Day 3 of Husker AI Days brought a fundamentally different tone than the Google sessions the day before. Where Google showed what AI can generate, MathWorks showed what AI must survive. The through-line was safety-critical, regulated systems: medical devices, power grids, aircraft, autonomous vehicles. AI is transforming all of these, and MathWorks' position is that the question is not whether AI gets used in engineering and science, but whether it is used responsibly, predictably, and verifiably. The day moved from that framing through a live MATLAB Copilot demo, a hands-on low-code workshop where participants built clustering, classification, and deep learning models without writing code, and closed with a deep dive on functional safety standards and digital twins for predictive maintenance. For Prairie, the day sharpened the connection between AI tools and engineering fundamentals, and gave the curriculum team a concrete picture of what "Problem Solving with AI" can look like when the stakes are real.

What Happened

The day opened early by Husker AI Days standards, with coffee at 8:30 and introductions in Kiewit A251. Gen Sasaki grounded the session in MathWorks' identity: a company that has been building AI toolboxes since 1990, focused on engineered systems in safety-critical, regulated environments. This is not consumer AI. The question she posed to the room was not whether AI will reshape engineering careers, but whether graduates will know how to use it responsibly. She ran a quick audience assessment: a mix of students and faculty, most familiar with MATLAB, fewer with Simulink, and almost no one aware of MathWorks' AI capabilities. That shaped the rest of the day.

Armando Garcia, a UNL graduate and MathWorks customer success engineer with 10 years of industry experience, took over for the GenAI tools demo. He walked through MATLAB Copilot (chat panel, inline code suggestions, automated comment and test generation), then previewed Simulink Co-Pilot and a new AI Tutor, both launching with the 2026A release within days. The demo that drew the most attention was the MCP Core Server, MathWorks' open-source toolkit for connecting AI agents to MATLAB, built on the same Model Context Protocol that Anthropic developed and that GitHub Copilot and Gemini CLI now support.

The room moved to A549 for the Low-Code AI Workshop, the most participatory session of the day. Nirav Acharya led participants through three exercises using human activity recognition data from accelerometers and gyroscopes: unsupervised clustering with k-means, supervised classification using the Classification Learner app (12 models trained simultaneously without writing code), and deep learning with a bidirectional LSTM neural network in the Deep Network Designer. The neural network reached 87% accuracy on test data, outperforming the traditional ML models at 76-80%. The session closed with a live deployment demo: a trained model running on an Arduino with a real accelerometer, classifying human activity in real time.

After lunch, the afternoon split between AI with Signals and Images and AI in Safety-Critical Systems. The safety session was the intellectual anchor of the day. It covered functional safety standards (IEC 61508, ISO 26262, DO-178), walked through the W-model for AI-augmented product development, and presented a DO-178 DAL-D case study on runway sign classification for aircraft. The live demo combined a physics-based Simulink model of a pump with machine learning for predictive maintenance, using the digital twin to generate synthetic training data and a classifier to detect fault modes (blockage, bearing wear, seal leaks). The key message: AI in higher safety-critical applications is experimental and not in production. Every safety-critical industry is testing it, but crossing from prototype to deployment requires validation that does not yet exist for probabilistic models.

The day wrapped at 3:30 with a debrief and agreement to reconvene for the May 11-12 MATLAB Workshop.

Schedule

The Sessions