Presented by Intel
COVID-19. Hurricanes. Power outages. Even without these extraordinary challenges, the last two years have offered plenty of new pressures to solve complex problems and do things differently. For heads-up enterprises, calamity has joined necessity as the mother of invention.
“In 2020 and 2021, the world had to pivot quickly to overcome big challenges across many fronts,” says Sandra Rivera, executive vice president and general manager of the Datacenter and AI Group at Intel. “To respond effectively, infrastructure needs to be flexible, scalable, highly secure and optimized for real-time data analytics and machine learning. Achieving that in an agile and cost-effective way requires best-in class platform solutions with tightly coupled hardware and software.”
Here are four real examples of diverse challenges solved by large-scale digital transformation using large data sets and flexible infrastructure. What do they have in common? Growing volumes of data, critical requirements for accelerated processing to handle the intensity of AI/ML algorithms, heightened security concerns, and the need for reliability that may exceed traditional mission-critical systems.
Supply chain: Smarter ways to optimize order fulfillment
From potato chips to semiconductor chips, disruptions in supply chains caused by the pandemic and natural disasters continue to wreak havoc with order fulfillment. Consumers and businesses alike have been impacted; with manufacturing, construction, and retail hit hardest.
“Everybody’s buying groceries online,” says Rivera. “If your milk doesn’t show up when you want it, that’s a problem. You want to get that kind of insight early, so the delivery truck can ensure the milk arrives as scheduled.” Such everyday problems pale next to the massive, ongoing disruptions faced by auto companies, vaccine makers, and other businesses worldwide.
Against this backdrop, a team of leading industry partners including Intel, SAP, and Red Hat has developed a novel new approach to optimizing order fulfillment. The modern supply chain system and order management solution uses advanced analytics and machine learning to proactively predict and prescribe options that can help avoid delivery delays.
Here’s how it works:
- An incoming order triggers two ML models based on historical order delivery data.
- One model predicts the likelihood of delay, and multiplies it by the revenue (or margin) of the incoming order to yield the expected revenue (or margin) at risk.
- Another model predicts the expected length of the delay, in days.
- The order is prioritized for fulfillment based on expected revenue at risk and expected delivery delays.
Calculating an expected delivery delay and the revenue margin (or risk). Credit: Intel and SAP
An executive dashboard delivers the predictive delivery risk score and a “days-delayed” forecast estimate. This knowledge empowers order-fulfillment and distribution center supervisors with data-driven insights and enables them to take corrective action prior to an order-delivery delay.
How they built it
A wide range of data sources feed the AI/ML models, including CRM, Material Master object model, sales and distribution monitor, Retail industry solution, and SAP S/4HANA Cloud. The variety and integration make it easy to consider inventory/out-of-stock data, routing information, and specialized handling requirements like climate control and customs/regulatory controls, and many others.
The smart system is notable because it also can tap into relevant third-party data sources, such as international customs and trade regulations, weather, and traffic. It can also use social-media-driven insights into trending product shortages and first-responder containment strategies and needs. These capabilities can help prepare and empower enterprises to proactively manage their supply chains — no matter what the operating environment calls for.
Core supply chain functionality is provided by The SAP Data Warehouse Cloud and on-premise SAP Data Intelligence. Generation Intel Xeon Scalable processors and extensions help handle the high volume of data and need for fast insights. The Red Hat OpenShift Container Platform and Storage serve as key building blocks. Integration is by Inspired Intellect, an end-to-end service provider of data management, analytics, and application development.
Developers used the Intel oneAPI Deep Neural Network Library (Intel oneDNN) to improve the productivity and performance of the deep learning (DL) framework. This included taking advantage of Intel Xeon’s built-in AI-acceleration capabilities, like bfloat16 support, to speed up training for AI models. The Intel DAAL library optimizes performance and scalability. Automated capabilities of SAP Data Intelligence help simplify development of ML algorithms.
Benefits for business and technology groups and customers
For technology groups, the new approach delivers enterprise-grade scalability in performance and DevOps. For business, the new system helps boost recognition of potential revenues. It also helps mitigate the risk and impact from recurring defaults on service-level agreements (SLAs).
Use is not limited to supply-chains. According to Rivera, the reusable and scalable framework can be applied to use cases and business challenges in a variety of verticals. Most importantly, she says, the new approach helps improve customer experience and satisfaction.
Bottom line: Holistic, data-driven decisions let enterprises better anticipate and respond to the dynamic demands challenging stable supply chains.
Utilities: Drones with data intelligence for safety measures
2021 continues the scorching march of wildfires worldwide with 43,555 so far in the U.S. alone. Utilities in California and elsewhere struggle to service far-flung poles, lines, and other electrical equipment that can spark catastrophic blazes and power outages.
A new co-innovation by SAP and Intel uses unmanned aerial vehicles with data intelligence to deliver a birds-eye view of potential utility problems and schedule preventive maintenance. The drones capture real-time image data of wiring, transformers, and other infrastructure feeding transmission and distribution lines. That makes it far easier to, for example, quickly spot plastic debris stuck in wires, a rotted wooden pole, or a tank located too close to other equipment.
High-risk areas are identified and prioritized by deep learning algorithms in the Intel Geospatial platform, an open, cloud-based platform for multisource geovisual data management and AI-powered analytics. Experts at the utility then can log into the cloud system for a top-down view of the findings, which are color-coded by equipment type and risk level. Once the inspection analysis is complete, results are pushed to SAP S/4HANA plant maintenance. This triggers notifications, which are bundled into field work orders using the SAP Asset Manager mobile app to address anything that requires attention or human intervention.
Visual field data from aerial drones triggers AI-guided notifications and work orders for utility crews. Credit: Intel and SAP.
Having UAVs collect intelligent data that connects back to equipment inspection, scheduling, and maintenance could be the future of automated wildfire risk mitigation, says Ron Gray, senior solution engineer at SAP.
“With analytical insights on where the biggest potential hazards are, electric companies can develop a prioritized schedule of inspections and maintenance plans, including outage management timeframes,” Gray says. “This would also help utilities correct missing or inaccurate information on equipment with fact-based mapping data, and prove compliance with regulatory reporting mandates.”
Go deeper: AI-Powered Visual Inspection (Video)
Healthcare: COVID management and safe return to work
When COVID-19 hit, Mercy Healthcare, like many providers, needed to quickly answer critical questions about how to ensure worker safety, handle capacity planning, and deliver the best patient care.
Administrators at the St. Louis-based regional system soon realized that getting the best answers depended on better data sets. “You can’t answer in-depth questions from higher-level data sources,” says Curtis Dudley, VP of enterprise analytics and data services for Mercy. “The key to creating better datasets that lead to insight is to dig deeper into the unstructured data.”
So, Mercy decided on a novel approach: repurposing an existing analytics system to combine structured and unstructured data to create a smarter view of how COVID-19 was spreading and how it affected patients at its 51 facilities.
Teams blended structured data from its EHR and other data sources with raw, unstructured data from clinical notes. Working with Intel and SAP, researchers used NLP and AI to develop a new way to transform the notes into structured, indexed data usable by the analytics software. (Data was anonymized to avoid matching with a specific patient.)
At Mercy Healthcare, NLP and AI turn more than 700 million clinical notes into structured, indexed data usable by analytics software to improve treatment of COVID patients. Credit: Intel
The combined data gets loaded into an SAP HANA database running on servers equipped with Intel Xeon Scalable processors, Intel Optane persistent memory (PMem), and Intel Ethernet 700 Series Network Adapters. Finally, researchers created analytics models and fed the results into custom dashboards and reports. In building the analytics system, Mercy digitized, curated, and enriched more than 700 million clinical notes.
Dudley says the system gives researchers a much deeper view of how medications, medical devices, capacity data, and treatments affect patient outcomes. Ultimately, it provides a better understanding of how to treat COVID-19 patients.
“This real-world evidence (RWE) platform is helping Mercy and some of its neighboring health systems, in addition to drug and medical device makers, address the needs of patients and healthcare providers,” he says. “It will continue to provide value after the COVID-19 pandemic recedes.”
Calculating and managing COVID risk while preserving privacy
“What happens when everybody starts coming back?” Heather Morrison, head of the SAP Global Marketing Experience Center and others began wondering last summer. How can customers and employees at these innovation centers worldwide feel and be protected and safe beyond wearing masks and practicing social hygiene?
Today, that anything-but simple challenge takes on new immediacy, as organizations around the world grapple with how to move from virtual-only meetings and safely bring people back to the office. A new pandemic cohort management dashboard helps organizations manage personal and collective risks from people returning to work as the pandemic rages on.
To do so, the pilot project gathers data on attendees before, during, and after in-person meetings, explains Kai Wussow, head of digital transformation at SAP. People are asked; “What have you been doing in the last two weeks? Who have you met? Have you been to a concert or a religious service? To a school or business meeting?” The data is digitized and used to assess the risk of in-person meeting — and makes it easier to alert public health authorities in case of an infection or outbreak.
From the outset, data privacy and security were of paramount importance. Europe’s strict GDPR rules and other regulations make it crucial to protect private data. Says Wussow: “Data needed to be secure and available. We needed to store little bits of information as safely as possible.”
The Pandemic Cohort Management Suite gathers and shares data from citizens, destinations, and government health departments to support both prevention and mitigation measures during pandemics. Credit: Intel and SAP
In a typical group-event scenario, each individual attendee downloads a mobile app onto a smartphone that connects to a virtual machine (VM) with Intel SGX Safeguard Extensions.
Intel SGX creates security-enabled enclaves in memory where data and application code are protected. The app serves as a virtual “pandemic wallet.” Like a real wallet, it holds valuable, encrypted personal data that only the individual can view (including name, address, testing, tracing, vaccination credentials, and current health symptoms). Its main function is to calculate the person’s risk of being infectious or becoming infected during specific planned activities. Risk is assessed both on medical data as well as activities during the past two weeks. Both are stored in the personal pandemic-diary section of the app.
Normally, after collection this data would be encrypted and sent to servers, where it is unencrypted. In the reinvented approach, data is kept in the secure, on-chip enclaves on the server CPU where collective risk is calculated; encrypted data is never shared. The net effect is end-to-end security and privacy for individual and collective data.
The project impressed officials in Heidelberg, Germany, who launched a proof-of-concept project. “We can safely collect different data depending on whether the person is going to, say, a restaurant, swimming pool, or university,” says Mayor Eckart Wurzner. “The ability to send information to the health department with privacy is one of the leading points.” He concludes: “This approach is future-oriented and innovative. It’s exactly what we need.”
As workers return to work, developers say the dashboard can play a key role in slowing COVID. Says SAP’s Morrison: “Because all the facts are digitized, you can almost automatically isolate the people who need to be isolated, manage the risk, and break the chain of infections.”
Whether sparked by crisis, opportunity, or a combination of both, digital transformation using proven technology and providers gives enterprises the flexibility to securely and quickly innovate smart, mission-critical applications in an uncertain, fast-changing world.
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact [email protected].