Saying yes and constantly learning: A recipe for adventure

The journey that has led me to Pontem

I grew up in a middle class South Indian family mostly reading The Hardy Boys, Enid Blyton (surprisingly not popular in the US, I now know) and Agatha Christie books. I was also a geek who read Dorling Kindersley encyclopedia books cover-to-cover and participating in trivia competitions. I fantasized about seeking adventures and I loved learning things. While I didn’t solve mysteries or find hidden treasures, I’ve nurtured the instinct of saying yes to the unknown in my career and life in general.

After high school, I wanted to experience another part of India, so I decided to join a university that had only just been formed in a part of the country where I hadn’t been before, with a language I didn’t understand, to major in Petroleum Engineering, a subject I knew almost nothing about! During an eventful 4 years at Pandit Deendayal Energy University, I had the opportunity to visit The University of Tulsa for a summer program. I visited the historic North Campus research facility and took classes with some world-renowned professors. I was fascinated at the scale, quality and applicability of their research.

Me after I visited the North Campus at TU

After I graduated, I decided to apply to TU to see if they would have me and luckily they did! I joined for my MS in Petroleum Engineering under Dr. Eduardo Pereyra and Dr. Cem Sarica. It was a quick and steep learning curve where I had to design an experimental facility (which our hard-working technicians had to actually build), understand how pressure transmitters and flowmeters worked, design and run experiments and make sure nothing blew up. I studied the use of downhole foamers and venturi nozzles as an artificial lift technique and even got to do a little bit of CFD. Those two years were a blast and when my advisors asked me if I wanted to stay around for a PhD, I immediately said yes.

For my PhD, I joined the recently formed Tulsa University Horizontal Well Artificial Lift Projects (TUHWALP) and got the opportunity to work on a behemoth of a test facility. This was a 250 ft long, 6 in. ID horizontal well test facility which could simulate both toe-up and toe-down horizontal well configurations. I studied the application of gas lift in horizontal wells, as well as the onset of severe slugging in toe-down horizontal wells. Take the madness from my Masters project, and double it! But it was an experience I cherish and I got to graduate at the end with a PhD after four years.

Thankfully I never had to go up this high, usually.

Image courtesy: University of Tulsa, Horizontal Well Artificial Lift Projects

Shortly after I finished my PhD, I decided it was time for another adventure and when Wood came calling, I decided to move to Houston in 2018. I spent the first couple of years on flow assurance projects for clients across the world. Again, another steep learning curve not just in technical aspects, but also in terms of managing clients, software licenses, managing expectations and delivering results. While there, I met Denis Akhiyarov who was super interested in data science and AI. I had the chance to sit in on a few of his presentations and my interest was piqued, but I didn’t know what to do with it or where to begin.

In the fall of 2019, I read a book called Football Hackers: The Science and Art of a Data Revolution by Christoph Biermann about how the use of data is revolutionizing the game of soccer. Now, I was hooked and I sensed the beginning of a new chapter. I joined an online data science bootcamp and learnt the ropes of working with data and creating useful models. I even created my own expected goals (xG) model as part of this program.

Using xG to tell the story of a game

Then a couple of major things happened. First, the pandemic struck and we all moved to working from home. Second, a new wave of industrial digitalization was on the horizon and opportunities emerged which were a confluence of process engineering and AI. Since I was already finishing up a program on data science, my manager at the time (shout-out to Shirley Ike) asked if I was interested in projects which were along those lines. Again, I immediately said yes!

What unfolded over the next 3 years were arguably the most enjoyable in my hitherto short career.

I was heavily involved in Microsoft Bonsai projects which involved training and deploying deep reinforcement learning agents to handle industrial use cases from process control to shop scheduling. I expanded my horizons beyond the energy sector and worked on use cases in fertilizer manufacturing, logistics, paint manufacturing and chemicals. It was an amazing experience and I learnt so much in such little time on the job. I was leading projects involving multiple stakeholders in multiple time zones, while learning about deep reinforcement learning on the side (I will leave the technical aspects of DRL and its applications in industrial use cases for another post). Microsoft however, unfortunately discontinued Bonsai in early 2023. In order to support a few projects in the short term, I led a crack-team of DRL practitioners and we developed an internal alternative using open-source libraries and AzureML within a few months.

The Deep Reinforcement Learning Training Cycle

Image courtesy: https://intellabs.github.io/coach

Working with Microsoft Bonsai, and developing our own alternative workflow was an illuminating experience. I learnt the value that process engineers and simulation experts who understand working with data and developing models can bring to industries that are going through waves of digitalization. I also learnt the value of understanding the best programming, deployment and maintenance practices aka MLOps. While the hype cycle around AI might wax and wane, the application of AI for industrial use cases to control physical processes is one that is more relevant than ever, from the stand-points of safety and productivity. I have seen first-hand the value it can bring to operators who have to look at one less screen or have to guesstimate one less variable. At the same time, these technologies should be applied responsibly, after rigorous testing and developed through reproducible, transparent and explainable workflows and has been my area of focus recently.

When the brilliant team at Pontem gave me the opportunity to lead the team in delivering and deploying AI solutions, I also immediately said yes! I am grateful for the experiences and opportunities I’ve been provided which I said yes to and the support I received every step along the way both from colleagues and my family. I am incredibly excited for this new chapter and can’t wait to get started. I hope to share more soon!