P-Rajasekaran

Post-COVID will be more focused on contextual data science

It is trying to fill a white space in the renewable space that not many have addressed so far. It manages over 5 GW of energy assets across the globe already. So what is it exactly that a sharper eye – or shall we say AIye – can do for a wind or solar energy park? P Rajasekaran, Chief Products Officer, Bahwan Cybertek helps us understand that while also covering some old and new challenges that surround Analytics. Excerpts from an interview:

DQ: Tell us something about the genesis of retina360. How is it different from other analytics tools in the market already?

P Rajasekaran: retina360 is an advanced AI platform, which leverages AI, ML, and Deep Learning algorithms to provide a 360-degree view of assets. Using retina’s real-time augmented analytics capabilities, many asset-intensive companies are able to reduce O&M costs, streamline operations and improve efficiency.

We started off by developing multiple solutions for asset-intensive industries. Exposure to diverse industries helped us spot the white space in the renewable sector. We realised analytics could help the wind, solar, and thermal park owners predict maintenance and save downtimes and costs. And we invested time, effort, and resources to build a predictive analytics solution exclusive to the renewable sector. Now, retina360 is a comprehensive predictive intelligence platform that manages over 5 GW of energy assets across the globe.

Extensive domain knowledge is our unique differentiator; it helps us build features like power curve analysis and yaw misalignment detection, specific to the wind energy industry. You need a good knowledge of renewable and a thorough understanding of the pain points of power producers to remain successful in this niche domain. Today, we are one of the very few analytics solutions providers from India that caters to wind, solar, and thermal energy producers in India and abroad.

You need a good knowledge of renewables and a thorough understanding of the pain points of power producers to remain successful in this niche domain.

DQ: What outcomes can asset-intensive enterprises and process industries tap here? Any specific examples of success stories you can cite?

P Rajasekaran: The common minimum denominator of the outcomes that are normally seen is productivity improvement or improved asset and process operations, leading to improved profitability. retina360 has helped our renewable customers minimise asset downtime of wind turbines and improve the performance efficiency of wind turbines and solar panels. Allow me to explain through a few use cases.

One of our global customers, who manages close to 1GW assets, detected damages on 26 wind turbines and the 1.7% drop in power production with the solution developed on the retina platform. The solution helped the customer to expedite turbine service and prevent production loss.

Another customer of ours identified 50+ critical turbine faults in FY2020-21 using the retina platform. These include generator failures that lead to extended downtimes. Preventive maintenance helped the customer to improve the productivity of their wind parks.

A global customer detected damages on 26 wind turbines and the 1.7% drop in power production with the solution developed on the retina platform.

DQ: Would it easily tie back to legacy environments, especially when we consider plants and process industries in India?

P Rajasekaran: Yes, retina360 comes bundled with a few built-in data integrators from very old and legacy systems. If there are proprietary data sources, we would create data integrators to access the data and ingest the same into retina360 for its appropriate usage.

DQ: Can you explain more about evolutionary algorithms and the role of a patent that will play out here?

P Rajasekaran: There is a classification called evolutionary algorithms in the area of optimisation and data science. The notable algorithmic techniques are Genetic Algorithms and Genetic Programming. However, if I can assume that the question is about evolutionary algorithms in the context of a patent, then I think the point here is that what is patented is an algorithm or a collection of the same that results in a specific outcome. If the algorithms were to be improved upon or altered in the future, based on several factors including technological improvements, market needs, etc., then the patent can be extended as per the provisions of USPTO. The patent would help protect the invention as well as secure the sole right to claim the benefits provided to the user using the patented algorithm.

Analytics applications are slowly becoming ‘AI aware’ and are beginning to provide not just hindsight but also foresight into various areas of usage.

retina360 recently patented its intelligent decision synchronization technology. The ‘Intelligent decision synchronisation’ USPTO patent defines the capability of the retina platform to use historical data, correlate to the business context and provide real-time insights to industrial users. The platform can combine data from multiple sources, build analytical models to detect potential equipment failures, and recommend specific actions that can prevent unexpected downtimes.

DQ: How much, and how soon, would data stories, decision intelligence, and photonic computing get mainstream in the field of analytics?

P Rajasekaran: I think these would be mainstream in this decade. More and more applications already existing would get intelligent and more unrealised use cases would become possible to be converted into really sophisticated and advanced applications or tools with high levels of usefulness.

DQ: Do you agree that almost every analytics solution is backward-looking and works on quantitative muscle? That may not always work for getting a good insight into the future which is often random, unpredictable, and powered by disruptors?

P Rajasekaran: Yes, I agree to a large extent. However, with more AI and ML becoming mainstream, this scenario is fast changing. Analytics applications are slowly becoming ‘AI aware’ and are beginning to provide not just hindsight but also foresight into various areas of usage. General-purpose applications of Google such as Google Now or even Google Maps show future states based on dynamically changing scenarios.

DQ: In the same vein, what would change in the post-COVID world? Can businesses still work on data gathered in the pre-COVID world?

P Rajasekaran: It depends on the source of data. If we limit the scope of data aggregation only to assets and machines, then using pre-COVID data will not change the outcomes. At the most, calibration of data points may be needed. But if you consider algorithms that combine data from people, you are facing a challenge. However, with the natural trajectory of data science heading towards small and wide data to power analytics, AI will become less data-hungry and more contextual. Gartner predicts that by 2025, 65-70% of organisations will shift from big datasets too small and wide data.

With the natural trajectory of data science heading towards small and wide data to power analytics, AI will become less data-hungry and more contextual.

But the sentiments of people have changed post-COVID. Industries like banking and healthcare cannot afford to miss the people context in data aggregation. In either case, data is vital to businesses. It would be wise to say that post-COVID will be more focused on contextual data science as most businesses run digital today.

P Rajasekaran IS Chief Products Officer, Bahwan CyberTek

By Pratima Harigunani

Leave a Reply

Your email address will not be published. Required fields are marked *