Topics In Demand
Notification
New

No notification found.

Edge – A key enabler of AI-as-a-Service (AIaaS)
Edge – A key enabler of AI-as-a-Service (AIaaS)

1412

0

High cost of AI-specific hardware, storage and compute capabilities have been among the top hindrances in large-scale adoption of AI/ML technologies, particularly for smaller organizations who understand the value AI can bring to their operations but are skeptical about the form and shape their investments would take. New on-demand delivery models, being termed as ‘as-a-service’ models, have emerged that can significantly help increase the adoption footprint of AI across a range of industries. These solutions are cost-effective, easier to implement, and can be refined continuously for better outcomes. An effective enabler is the case and potential for customization even if use-cases are the same across industries. These innovations are riding the wave of adoption owing to:

 

  • Low cost of infrastructure, particularly Storage and Compute
  • Access to large volumes of data
  • Democratization of ML, easy availability of Open-Source tools and platforms

It may seem that the new AI delivery models mirror the existing Cloud Computing models – comprised of SaaS (on-demand software), PaaS (development, integration etc.) and IaaS, but there are other elements that fall under the scope of AIaaS (AI-as-service). While some approaches do adhere to this model, this space is still fluid and evolving.

 

In this article we discuss the role of the trio of Connected Devices (AI-IOT), Cloud and Machine Learning in the delivery of AI services

 

AI solutions leveraging Microservices and APIs

 

Generation of insights require large compute capabilities, which can be made easy with cloud computing. Some of the evolving trends in this space are:

  • Servitization of Artificial Intelligence, that has come about with low compute costs
    • Move towards ‘Outcome-as-a-service’, rather than one-off instance of sale or delivery
  • New delivery models (on the Edge) are emerging which have unique go-to-market and pricing strategies. One such example is ‘Inference-as-a-service’.

While AI/ML solutions can be delivered over the cloud, there are certain industries where data privacy, access and latency are key decision factors on the choice of delivery models. For example, in mission-critical applications in the Healthcare, Manufacturing, and Power sectors, Cloud can often be a disadvantage where real-time image quality, inferences/decisions can directly impact the outcome of investments. It can be safely said, that Microservices architecture is the go-to methodology for scaling AI applications.

 

What is AI on the EDGE?

Today, interconnected systems/devices make it possible to deliver services locally on the device or on the EDGE, as it is popularly known. Improvements in communication infrastructure (5G/low-latency networks) and Neural Computing is pushing the adoption of EDGE computing further. Compute capabilities exist on devices which make for easier adoption. This has a direct effect on companies which are trying to implement AI solutions, creating incentives for fence-sitters to identify use-cases and apply AI to solve some of the problems.

 

“In this model, the data is stored on the cloud where the algorithms/models are built and run but are deployed on the Edge. This can take the form of ‘Subscription-as-a-service’, or ‘Inference-as-a-service’. There are industries/use-cases that require AI-powered solutions closest to real-time.”

Use-cases in select industries

  1. Industrial and Manufacturing domain: AI on the EDGE can help solve a multitude of use-cases in this sector. AI delivered using microservices architecture on the device can open-up new vistas of service delivery where the customer is charged based on a set of pre-decided positive outcomes.
    1. One of the most important use-cases is Predictive Asset Maintenance. Assets have natural downtime & maintenance schedules as well as ‘Service Hours/Life’. Thus, it is important to rationalize the use of machines. As the name suggests, Predictive modeling techniques (using regression models) are used to monitor potential failures, reduce the frequency of maintenance, and improve asset longevity. The idea is to assess the duration of operation uptime before it requires maintenance. It can also be used to reliably predict remaining useful life of the asset.
    2. Anomaly Detection: Anomaly detection often includes video surveillance for assets which are spread out across physical sites.
    3. Defect Detection is a big problem in manufacturing domains such as steel and automotive. The evolution of as-a-service models can help organizations to identify needs and map it to AI – vision, audio, or real-time streaming problem.
  2. OTT/Streaming Media: AI is poised to transform the global OTT space. The stiff competition among the global OTT majors to create competitive differentiation in Content creation and delivery has opened-up multiple possibilities. Providers understand that AI is a key input to gain competitive advantage. Improving user experience and viewer engagement (personalization/recommendation engines), rationalizing the cost of transmission, and improving rendering ratios (VST and VRT) are some of the use-cases that can be solved using AI/ML technologies. What AIaaS does is provide flexible pricing options linked to a cost-saving outcome, which can result in higher adoption of AI in this segment.
  3. Power/Energy: Deploying AI on the EDGE can help asset-heavy companies in the sector to plan for maintenance downtime and act proactively in predicting asset failures before they occur. This is particularly true for assets which are dispersed over large geographical areas where access can be challenging at times. This can have a positive impact on lowering maintenance/overhead costs and avoid service disruptions.

Scoping out the contours of evolving AI delivery models is among the top focus areas of a study NASSCOM is conducting. It would also be interesting to analyze the interplay and overlap of the various elements within the provider/adopter landscape. In this article we have discussed how Edge can be leveraged to expand key use-cases of AI in select industries. In the forthcoming articles, we would cover other emerging consumption models of AI and their application across industries. So, stay tuned..


That the contents of third-party articles/blogs published here on the website, and the interpretation of all information in the article/blogs such as data, maps, numbers, opinions etc. displayed in the article/blogs and views or the opinions expressed within the content are solely of the author's; and do not reflect the opinions and beliefs of NASSCOM or its affiliates in any manner. NASSCOM does not take any liability w.r.t. content in any manner and will not be liable in any manner whatsoever for any kind of liability arising out of any act, error or omission. The contents of third-party article/blogs published, are provided solely as convenience; and the presence of these articles/blogs should not, under any circumstances, be considered as an endorsement of the contents by NASSCOM in any manner; and if you chose to access these articles/blogs , you do so at your own risk.


images
Bandev Ghosh
Senior Manager

© Copyright nasscom. All Rights Reserved.