Building Data-Centric Edge AI for Medical Device Predictive Maintenance
Online Training Workshop: Build, Learn, Deploy
Overview
This workshop is designed to introduce developers to the hardware and software required to build data-centric Edge AI applications for medical device monitoring and predictive maintenance. Participants learn how medical data, such as blood-glucose level and heart rate, is acquired, processed, and managed deterministically at the device, and how this data is transformed into AI-ready features for on-device inference. The workshop covers the selection and integration of edge hardware (MCUs, MPUs, NPUs, sensors, and accelerators) alongside production-grade software for real-time control, time-series data management, signal processing, and Edge AI frameworks. Through hands-on exercises, developers gain a clear understanding of how data management, analytics, and AI work together to detect anomalies, predict failures, and enable reliable, explainable motor health intelligence directly at the edge.
Workshop Date: Thursday, June 4th, 2026
Workshop Time: 11:00 AM–2:00 PM PT (2:00 PM–5:00PM ET)
Workshop Location: Online
Online Tutorial Workshop Syllabus
Objective: Detect abnormal conditions in real time on an Edge device for medical systems. Build data-driven firmware to ingest physiological and device signals, generate cleaned features, apply Edge AI for anomaly detection and predictive maintenance, and visualize system health and patient/device status over time.
The workshop comprises three 60-minute modules.
Module 1: Data-Driven Firmware Design and Programming
- Sensors for medical and device monitoring
- Sample signals such as ECG, heart rate, respiration, temperature, pressure, flow, and device telemetry
- Hardware setup
- Embedded device, storage media, and communication interfaces for medical applications
- Data modeling for edge devices: patient monitoring & predictive maintenance
- Ingest and store sensor samples, features, inference results, events, and system metrics
- Define sampling rates, stream window sizes, and key data structures
- Data cleaning and conditioning with deterministic stream buffering
- Per-signal filtering, normalization, artifact removal, outlier detection, and handling of missing data
- Transactional commits and time-series storage for reliable and safe data management
- Feature engineering for Edge AI
- Extract features such as mean, variance, peaks, trends, and domain-specific indicators (e.g., heart rate variability, signal morphology, flow patterns)
Module 2: Building an Edge AI Pipeline
- Select AI model algorithm
- Anomaly detection vs. classification for physiological signals and device behavior
- Datasets with variations and disturbances
- Physiological variability, sensor artifacts, abnormal vital patterns, and device performance deviations
- Train a model with an Edge AI development tool
- Train using normal physiological and device conditions, validate with abnormal scenarios
- Assess model size, latency, and deterministic performance
- Export optimized model to embedded runtime
- Real-time inference and monitoring pipeline
- Integrate ML inference into feature streams to generate anomaly scores or classifications per time window
- Define thresholds, persistence logic, and system state models (e.g., normal, alert, critical)
- Inject simulated anomalies and observe model response and transitions
Module 3: Edge Device Analytics, Visualization, and Connectivity
- Communicate with ITTIA Data Connect over Ethernet, WiFi, and/or serial interfaces
- Create a custom on-device dashboard with ITTIA Analitica
- Display metrics such as vital signs trends, device performance indicators, anomaly counts, and system status
- Visualize data via web dashboards, web services (HTTP/REST), Python, and C/C++
- Analyze patterns and system behavior
- Frequency, duration, and severity of anomalies
- Trends in patient data and device performance
- Correlate multi-signal data
- Relate physiological signals with device telemetry
- Identify interactions between different measurements
- Score system and patient/device health
- Evaluate anomaly counts, severity trends, and performance indicators
- Generate real-time health or alert levels
- Use anomaly data for diagnostics and predictive maintenance
- Support early detection of patient risk or device degradation
- Enable proactive maintenance and intervention
- Validate system performance
- Ensure bounded latency per processing window
- Maintain efficient storage usage and long-term reliability
Workshop Sample Application
During this hands-on workshop, you will be trained to design, build, and deploy a real-world, data-driven AI application, gaining practical skills you can immediately apply to embedded and edge systems.
Anomaly Detection in Medical Devices
Participants take on the role of an embedded engineer responsible for ensuring reliable operation of a medical device in real-world conditions. Using MCU-based platforms, they capture live physiological and device signals directly at the edge, without reliance on cloud connectivity. An on-device AI model is trained to learn what “normal” looks like for both patient signals and device behavior, adapting to variations over time.
As the system operates, subtle anomalies begin to emerge, irregular heart rhythms, unexpected temperature changes, signal artifacts, or device performance deviations—early indicators that traditional threshold-based methods may miss. The AI detects these anomalies in real time, while ITTIA DB Lite AI persistently stores and organizes both data and inference results on the device.
By combining AI with deterministic data management, raw signals are transformed into actionable insights. Anomaly frequency, duration, and severity are analyzed to generate real-time health and device status scores, alongside on-device statistics such as vital trends, peak values, and anomaly counts.
By the end of the session, participants build a complete Edge AI solution demonstrating how data-centric design and on-device intelligence enable reliable, explainable, and real-time monitoring and predictive maintenance in medical devices, directly at the edge.
Course Details
Workshop Cost: USD $250 per attendee
The workshop fee includes:
- Comprehensive lecture materials
Workstation: This is a hands-on workshop. You must have an STM32H573I-DK device, a USB-C cable, and a Windows PC:
*Unlock 25% OFF your STM32H573I order when you register!
Edge Data AI Certification for Embedded Devices
Workshop attendees will receive ITTIA DB Platform certification. This certification validates hands-on expertise in using the ITTIA DB Platform for deterministic data management, on-device data processing, and Edge AI–ready data handling. It demonstrates and validates the ability to design reliable, power-fail-safe, and AI-enabled embedded systems using proven data architectures, reducing development risk, accelerating time to production, and enabling high-quality, scalable solutions.
Past Workshop Participants
ITTIA has delivered highly sought-after, hands-on workshops that attract engineers and innovation leaders from across multiple sectors. Our sessions are trusted by companies building real-world, data- and AI-driven embedded systems, and have been attended by teams from leading organizations committed to advancing intelligent edge technologies, including:
Abbott – Acuity Brands – Advances Energy – Applied materials – Avery Biomedical Devices, Inc. – Badger Meter – Cardios – Cat Wranglers
Cognosos Inc. – Continental AG – CPI – Declarative Futures – EDS – Emtech – Everspin – EXFO – Honeywell – Kellogg Northwestern
L3Harris – Magna International – Mayo clinic – Megavolt Labs – Milwaukee Tool – Mold Masters – Nidec Global Appliances – OTTO Engineering
Philips – Resideo – Schneider Electric – Siemens – Southland Sensing Ltd. – TITOMA – TTI Floor Care – Wind River – Zetron/Codan
Audience
These immersive, hands-on workshops are built for embedded architects and engineers who are shaping the next generation of data-driven, AI-enabled edge systems and want more than theory, they want results. Through practical, expert-led sessions, participants learn how to select the right device platform, design deterministic real-time firmware, and implement robust on-device data management that elevates data to a first-class system asset. The program goes further by showing how to prepare and integrate optimized AI pipelines directly into embedded workflows, while addressing cybersecurity, functional safety, diagnostics, and long-term lifecycle management from day one. By the end of the workshop, engineers leave with the confidence, skills, and real-world experience needed to build intelligent, trustworthy, and future-proof embedded systems ready for production.
These workshops are highly technical, intended for professional engineers and are not suitable for the general public without relevant qualifications.
