Satyajit Dwivedi

The AI Champion

Satyajit Dwivedi

Regional Director, EMEAP, Energy & Utilities, Manufacturing & Public Sector,

SAS

An electrical engineer with management in Accounts & Systems, Satyajit Dwivedi began his career 28 years ago as a trainee engineer performing substation and captive power plant maintenance in a soda ash plant. He quickly learned the difficulties of performing preventative maintenance on static and rotating equipment at a time when manual and paperbased processes were the norm. In 1993, he joined Hindustan Petroleum, where he worked on real-time data acquisition and control applications, such as PLC/SCADA systems, High Voltage AC systems, and control rooms. This was his first exposure to big data and that sparked his interest in the field of data analytics.
Post his management, he started his journey in analytics, developing mutual fund performance algorithms for stock portfolios that outperformed market indexes. During one of the government consulting engagements, he observed that most of the IT applications designed are open-loop systems. With exposure to big data and the need for closed-loop automated systems in the decision process, he realised that analytics would be the best space to work for in the coming many decades. In 2004, he joined SAS and has since provided immense value in the AI industry to customers across diverse industries and regions. TradeFlock interviewed Satyajit Dwivedi, who has 20 years of experience in AI and who worked as Director, Global Practice, Process Sensor and IOT, SAS to learn more about his journey.

Kindly give us a brief overview of your professional journey. How did you get involved in the AI industry?

When I started my career, few digital solutions were existing in accounting, business process automation, and plant control systems. The philosophy of data as an asset did not exist in corporations. When I joined IRIS after my management degree, my first work in analytics was when we developed an algorithm for evaluating fund performance and creating stock portfolios that consistently outperformed market index funds. Subsequently in 3i Infotech, when working on a couple of government consulting projects evaluating IT applications under the horizontal transfer program in 2003-2004, I noticed a glaring gap between outcome analysis and process feedback-based evaluations. Control systems theory was my favourite subject in engineering, and I believed a system would need continuous feedback for timely input and process corrections to operate in the optimal state. I concluded that analytics would be the best industry to work in over the next few decades. With that in mind, I joined SAS in 2004 and have worked in the AI industry ever since, delivering tremendous value to customers across numerous industries and regions.

Over the last three decades, what are the challenges you have overcome, and what learnings do you take from them?

I’ve had an exciting career journey, spanning diverse industries such as Chemical, Oil & Gas, Petrochemical, Capital Markets, Smart Government, IT Consulting & now AI powered Digital transformation. Working in a global role allowed me to work across different locations and industries, constantly adapting to change and developing value propositions that drive digital transformation initiatives with changing business perspectives involving profit, people, planet, and now purpose. One key factor that has helped me succeed in my roles is my ability to appreciate and adapt to Regional Director, EMEAP, Energy & Utilities, Manufacturing & Public Sector, SAS The AI Champion 10 Best Leaders from AI in India 2023 38 the specificities of each geographic area, most importantly the culture. Working in a global role in an emerging area in 2021 was quite challenging. This required very close engagement with regional teams, building regional capacity to deliver strong results for the organisation. To achieve my goals with confidence, I rely on self-motivation and constant learning. I firmly believe that having a high level of self-efficacy is crucial to overcoming challenges and turning them into opportunities for professional growth.

What is your roadmap? Can you please elaborate on some of your future strategies?

At any given point of time, my future strategy has been to unlearn, learn and relearn. The importance of sustainability and the need of rigour to every learning process and the spare time during COVID made me register for PhD in Water Science & Governance in TERI School of Advanced Studies. Future innovations should be aimed at mitigating the global risks that the world is currently facing. The greatest growth would come from AI-based solutions that address environmental and food security concerns, assist businesses in achieving net zero transitions, reduce carbon & water footprints, and provide equitable, cheaper, and faster remote access to healthcare. Sustainable living would be prioritised by companies driven by purpose rather than profit and AI should enable this journey. This is the best space to work for and my next two decades would be dedicated to younger generation pursuing “Zen and the art of earth maintenance”.

What are some of the most interesting projects you have undertaken? Can you please share your approach to risk management and modernisation?

In the last decade, I have designed and deployed many digital transformation programs powered by AI, both in pragmatic services and aesthetic services side. Challenges come with four things: a) understanding the business need and the relevance of AI to solve the business problem b) the granularity of the data and the history being considered, c) the type and the number of algorithms that need to be built for the solution, and d) integrating the whole decision process with real-time model deployment. Integrated planning is a core process in Oil & Gas downstream & Utility sector that involves demand-driven planning (long/medium/short/ very short time horizon) across the large network. There are several time series whose patterns need to be understood and future estimates arrived at with many influencing factors. In one of the O&G projects, 25000 demand time series were to be trained, forecasted daily, and integrated with supply planning. Different factors impact different time series to different degrees at different times. Deploying modern AI models for large time series requires both science and art. In a utility company, we undertook the modernisation of long-term forecasting and network capacity planning involving 1000 nodes in the distribution network. A significant increase in productivity and savings in cost was observed. Some of the AI-based predictive maintenance projects on rotary and static equipment for both diagnostics & prognostics have generated some astonishing insights, and the value per use case has been to the tune of $2–5 million per year. In the coming days, sustainability & ESG would drive digital transformation programs. I see digital transformations powered by AI would be focused on increasing energy efficiency, reducing carbon intensity, and optimising water footprint. Climate risk modeling and its integration with net-zero initiatives would become extremely important. Realtime monitoring of ESG metrics and developing scenario analysis to model the possible financial risks associated with climate change and assessing the resilience of individual financial institutions and the financial system AI would play a major role.

How do you approach the ethical considerations of AI development and implementation?

Although there has been no common acceptance of definitions of AI in the computer science and technology world, it is considered a constellation of many different technologies working together to enable machines to sense, comprehend, act, and learn with human-like levels of intelligence. Therefore, the deployment of any AI solution needs to carefully address both the epistemic factors and the normative factors. The epistemic factors include inconclusive, inscrutable, and misguided evidence, and the normative factors include unfair outcomes (algorithmic bias, AI hallucination), transformative effects, and traceability. Autonomous and semiautonomous systems with large historical data, numerous variables, and a mix of structured, and unstructured data, text, and images make the AI ethical consideration an arduous task. It requires human interventions to detect & adequately treat data quality issues, detect and transparently compensate for algorithmic bias, and gain the trust of the model, which is very important for continuous usage of the model. The model outcome is very dependent on data; therefore, careful application of the eight data quality dimensions is critical: resolution, accuracy, completeness, redundancy, readability, accessibility, and consistency. Any AI deployment should have data quality rigor. Second, AI systems are responsible systems built on complex relationships. Therefore, AI models need to be explainable to be responsible for addressing the six factors: clarity, compliance, confidence, consent and control, challenge, and continuous improvement.

Satyajit Dwivedi