[Editor's Note: Nuruddeen Inuwa Aminu is a member of the TWA Editorial Board and is the author of previous TWA articles.]
Utkarsh Sinha is a senior research engineer at Xecta Digital Labs whose work lies at the intersection of reservoir physics, data science, and practical energy innovation. With more than 8 years of experience, he develops physics-informed and hybrid machine-learning solutions for reservoir and production engineering, spanning reservoir management, enhanced oil recovery, artificial-lift optimization, production forecasting, and PVT/EOS modeling.
He has authored or coauthored more than 30 technical manuscripts, received over 260 citations, and is a co-inventor of six granted and pending patents. Alongside his industry role, he volunteers as a research collaborator with Birol Dindoruk, advancing hybrid frameworks that combine first-principles reservoir physics with modern deep-learning architectures.
Sinha is also an emerging leader within SPE. He served on the SPE Data Science and Engineering Analytics Technical Director’s Advisory Committee from 2020 to 2023, the PetroWiki Steering Committee from 2023 to 2025, and as secretary of the SPE Gulf Coast Section Young Professionals from 2021 to 2022. He has held technical and leadership roles at URTeC and the SPE Improved Oil Recovery Conference, including serving as the URTeC 2026 Theme Chair for Applications of Artificial Intelligence and Data Science. He also contributes to the profession as a technical reviewer for leading petroleum and energy journals, including SPE Journal.
He was recently named a 2026 TWA Energy Influencer.
He holds a BTech in chemical engineering from VIT University and an MEng in petroleum engineering from the University of Houston, where he received the BP Student Scholarship and the Engineering Dean’s Master’s Scholarship in recognition of his academic performance.
Nuruddeen Inuwa Aminu (NA): What first motivated you to study chemical engineering, and what made you pivot toward petroleum engineering for your master's?
Utkarsh Sinha (US): Engineering was almost a natural choice for me growing up. My father is a retired head of the electrical engineering department at a government polytechnic, and engineering was a common career path among my extended family and in the community around me. My earliest specific inspiration came from a cousin 4 years older than me, whom I greatly admired. He studied chemical engineering at an IIT and later worked with Shell in Bangalore and Houston on flow assurance. Following his journey showed me what the field could offer and helped me understand that chemical engineering is much broader than chemistry or laboratory work.
That initial exposure, and the interest it sparked in the oil and gas industry, deepened during my undergraduate internships, first at ONGC in source-rock characterization, and later at a research laboratory studying surfactant-oil interactions. Thermodynamics was already my favorite subject and discovering its application to reservoir fluid phase behavior and PVT was an immediate draw. Petroleum engineering, in turn, brought together my interests in numerical methods, fluid mechanics, transport phenomena, and porous-media flow, ultimately motivating me to pursue it for my master's.
NA: What inspired you to move from India to the US, and how has your personal and professional journey evolved during your years here? What was the transition like?
US: In my undergraduate chemical engineering cohort of about 40–50 students at VIT University, there were two common paths after graduation: accepting a campus placement in India or moving overseas to pursue a master’s degree. As I mentioned earlier, my growing interest in the oil and gas industry and its specialized applications made pursuing a master’s in petroleum engineering in the US a natural next step.
I took the GRE a little later than planned and missed the fall intake. During the gap, I briefly worked at Derrick Petroleum Services in India, supporting business and market analysis for the upstream oil and gas industry. I enjoyed the work and the colleagues, but even in those few months, I was confident about pursuing my master's, and to this day I consider it the best decision I've made. I was fortunate to have a cousin living in Houston whom I was close to, along with close extended family on the West Coast. Having people around you in a new country makes a real difference.
I made close friends that first semester. The industry terminology caught me off guard at first, but like everyone else, I caught up eventually. I learned to cook, found my footing, and by the end of the semester felt like I was hitting my stride. It was a period that shaped me, from a somewhat reckless kid into someone more independent and level-headed. The financial stakes were real too, coming from India with the currency conversion and cost of living, and that pushed me to become a more careful, deliberate decision-maker.
I finished that first semester strong and earned scholarships for the rest of my graduate degree, a relief that eased more than a little of the financial weight I'd been carrying. In my second semester, I took the PVT phase behavior class taught by Birol Dindoruk, a class I'd been looking forward to, and one made even more special by learning from someone who is a world-renowned subject-matter expert in the field. It was one of the courses I enjoyed most. The late nights in the computer lab on that project remain some of my best memories from that time—working hard, making friends, and genuinely enjoying the process.
NA: Can you tell us about Xecta Digital Labs, your role there, and what your experience has been like so far?
US: Xecta Digital Labs is a technology company that develops intelligent software solutions for the oil and gas industry. We combine traditional petroleum-engineering expertise with data science and AI to help energy companies better understand their reservoirs, improve well and asset performance, optimize production, and make faster, more informed operational decisions.
I have been with Xecta for over 4 years and currently work as a senior research engineer. The experience has been highly rewarding, as the work closely aligns with my interest and background in combining numerical engineering models with data-driven techniques to solve real-world problems. I have also benefited from exceptional mentorship. One of my former managers, Sathish Sankaran, who previously served as the company’s executive vice president and is highly respected across the industry, became one of the most influential mentors in my career. He took the time to rotate me across teams, entrusted me with some of our most challenging technical projects, and spent countless hours in one-on-one brainstorming sessions with me. These experiences broadened my expertise across reservoir and production engineering while strengthening my technical, leadership, and communication skills in ways that continue to shape how I approach my work today. More broadly, my colleagues at Xecta are technically strong, open minded, and always willing to engage in thoughtful discussions. I have learned immensely from both the fresh perspectives of younger colleagues and the deep expertise of senior technical professionals.
NA: What does your day-to-day work as a senior research engineer at Xecta Digital Labs look like, and how would you describe the domain you focus on?
US: I am part of Xecta’s research and development team, where my day-to-day work involves translating engineering ideas into algorithms and practical software solutions. This includes writing Python code, developing algorithms, validating models against simulated and field data, and collaborating with our product team to prepare successful prototypes for deployment. My primary domain is hybrid reservoir and production modeling, where we combine physics-based numerical models with data science. In some cases, a data-driven model improves a physics-based prediction; in others, it provides an informed starting point that helps the engineering model reach a solution faster. The objective is to develop tools that are fast, reliable, interpretable, and scalable. Once our models have been rigorously tested, they are transferred to the production team, which turns them into commercial software available to clients through Xecta’s subscription platform. My work has also contributed to 15 conference proceedings and six patents on which I am a co-inventor. It is always rewarding to share our work with the broader technical community, receive feedback, and see its potential impact beyond the company.
I began my time at Xecta on the unconventional-reservoir team, working on technologies for reservoir-performance forecasting, frac-hit detection and quantification, liquid-loading diagnosis, and artificial-lift selection and timing. Over the past 2–2.5 years, much of my work has focused on developing and managing reservoir graph network (RGNet). RGNet is Xecta’s hybrid, reduced-order reservoir model, combining reservoir physics with data science for rapid simulation, history matching, and production forecasting. By representing complex 3D reservoir systems as simplified networks of connected cells, it reduces simulation time from hours or days to some minutes while retaining the essential reservoir behavior. I also led a water to gas ratio forecasting project for a major operator’s aquifer driven condensate field, helping anticipate water breakthrough, forecast the subsequent rise in water production, and guide mitigation strategies. The project resulted in a 5-year commercial engagement and involved technical leadership, regular client communication, and successful solution delivery. My time at Xecta has allowed me to take ideas from early research through development and ultimately transform them into tools that create lasting value in the field.
NA: Looking back over the past 8 years, were there any pivotal moments early in your career that shaped your professional journey?
US: It is difficult to identify a single pivotal moment. Moving to the US, choosing the right academic advisor and building a collaboration that has continued for more than 9 years, learning from a talented team through technically challenging but rewarding work at Reach Production Solutions that provided the much needed technical foundation for my career, and later joining Xecta, where I deepened my expertise in hybrid modeling and helped develop emerging technologies with practical industry applications, have all shaped me in different ways. Each experience has added something valuable to my journey.
If I had to choose one, however, it would be moving to the US. It placed me in an unfamiliar environment, pushed me beyond my comfort zone, and introduced me to a world of possibilities where there was always room to learn and grow. It also led me to the mentors and opportunities that ultimately shaped my career.
Your first job is always special, and I was fortunate that my immediate manager, Rob Perry, was also the company’s CEO. His previous roles as VP of deepwater facilities at BP and director of global subsea processing reflected the depth of his knowledge and experience. Yet despite his distinguished career, he remained remarkably humble, approachable, friendly, and witty. His warmth made it easy to break the ice and learn from him. He taught me to stay grounded, continually strengthen my technical skills, and focus on creating value. Just as importantly, he showed me how to distill complex findings and communicate them clearly. Learning both these technical and interpersonal skills, and understanding their importance so early in my career, was invaluable.
These are some of the moments I remember most, but there have been many others. Ultimately, every experience and every person we encounter contributes in some way to who we become.
NA: For readers less familiar with the term, how would you explain "physics-informed machine learning" and why it matters for reservoir and production engineering?
US: Physics informed machine learning brings together two complementary strengths: the reliability of engineering physics and the speed and flexibility of machine learning. Traditional physics models may be limited by unavailable inputs or the computational cost of running detailed simulations and large sensitivity studies. A purely data driven model can be fast, but it may produce unrealistic results when it encounters conditions outside its training data. Physics acts as a guiding force, keeping the model connected to how the real system behaves.
That guidance can be introduced through physical constraints, prior engineering knowledge, physics embedded within the machine learning architecture, or data-driven reduced order representations of complex, high dimensional models. This is especially valuable in reservoir and production engineering, where field data are often limited, noisy, and constantly evolving. By blending physics with data, we can generate forecasts that are faster and more scalable while remaining realistic and interpretable.
The practical value is that engineers gain timely, actionable insights with numbers they can trust and confidently use to make decisions. Reservoir and production decisions often need to be made quickly across hundreds or thousands of wells, even when data is incomplete. Physics informed machine learning supports applications such as anomaly and root cause detection, well interference and liquid loading diagnosis, wellbore multiphase flow modeling, artificial lift and completion selection, injection rate and BHP control optimization, surface network modeling, and integrated asset management. It also helps compensate for the assumptions and subjectivity built into simplified physics models without sacrificing their engineering foundation.
Multiple models can be combined within automated pipelines that continuously process new field data and address several challenges simultaneously. This transforms periodic manual analysis into a continuously updating system that enables earlier intervention, improved production, and scalable decision support across an entire asset.
NA: Across the domains that you have worked on, which has been the most technically stubborn problem to crack with machine learning, and why?
US: Across the domains I have worked in, the most stubborn challenge has not been any single application. It has been making data science and analytics reliable when they move from a controlled development environment into the field. Achieving a good accuracy score on historical data is relatively straightforward. Building a model that remains physically realistic, interpretable, computationally efficient, deployment ready, compliant with operational constraints, and useful across thousands of wells with different geology, operating conditions, and data quality is much harder.
In reservoir and production engineering, the same observation can have several possible causes. A decline in production rate could result from something as simple as an increase in wellhead pressure, something more complex such as reservoir depletion, well interference, or liquid loading, or merely a faulty sensor. The model must distinguish among these causes rather than simply recognize a pattern. This challenge has been common across my work on interwell interference, liquid loading, production forecasting, artificial lift optimization, surface pipeline networks, material balance, and reservoir simulation. The solution is rarely a standalone data driven model. It requires combining engineering physics, operational context, scalable algorithms, careful handling of poor data and edge cases, rigorous field validation, and presenting the results in a form engineers can easily understand and use. To me, the real breakthrough is not when a model performs well in testing, but when engineers trust its results enough to use them in operational decisions on a frequent basis.
NA: With more than 30 manuscripts and 260+ citations, is there one paper or finding you're most proud of, or that changed how people in the field think?
US: That is a little like asking a parent to choose a favorite child. Each paper represents a different question, collaboration, and stage of my journey, so selecting just one is difficult. What gives me the greatest sense of fulfillment is their shared theme: combining physics with AI to solve practical engineering problems and then make those solutions accessible to others.
A few contributions are especially close to me. We developed models for predicting the minimum miscibility pressure of CO₂ and produced hydrocarbon gas, demonstrating how physics informed machine learning can reduce reliance on costly laboratory experiments and accelerate the evaluation of enhanced oil recovery, carbon capture utilization and storage, and gas reinjection projects. We also created a unified dead oil viscosity framework covering the full spectrum from very light to extremely heavy oils. In reservoir modeling, our research used physics informed neural networks and graph neural networks to create fast digital representations of complex reservoirs. These models preserve essential flow behavior and interactions among wells while significantly accelerating waterflood forecasting and scenario evaluation. Other studies produced scalable frameworks for quantifying interwell interference to guide well spacing and remediation, as well as diagnosing liquid loading across more than 1,000 gas wells.
One of our review papers mapped the rapidly evolving landscape of physics-informed machine learning (PIML) in subsurface engineering, connecting fundamental concepts and model architectures with practical applications, current limitations, and future opportunities. To make the paper’s content easier to explore and understand, I developed the PIML Chatbot, an AI research assistant that uses large language models and retrieval-augmented generation to answer questions, direct readers to relevant sections, and present complex information in a more accessible form. When released in early 2025, it was, to the best of our knowledge, the first research chatbot of its kind developed specifically for the oil and gas community. Through the IPB&F Consortium, led by principal investigator Birol Dindoruk, we also made several tools freely available to the community, including the PIML Chatbot and calculators for minimum miscibility pressure and oil viscosity, some of which were featured in The Way Ahead in 2023. Together, these projects reflect what I value most: research that advances science, solves practical engineering problems, and remains accessible to the wider community.
NA: You are a co-inventor on six patents, can you walk us through how one moved from research idea to real field application?
US: At Xecta, every patent began with a real field problem that lacked a reliable, practical, or scalable solution. We translated our domain knowledge into a novel algorithm and tested it rigorously using simulated and client provided data, which were often limited or imperfect. Once the idea demonstrated potential, we developed a proof of concept and evaluated it through focused client sprints with clearly defined deliverables. This process required much more than model accuracy; we had to consider operational constraints, value drivers, edge cases, missing data, and possible system failures. Regular client feedback helped us refine the technology and ensure its recommendations were understandable and actionable. When an algorithm proved original, impactful, and commercially scalable, Xecta protected the intellectual property through the patent process. Following client validation and acceptance, the solution was transferred to the production team, deployed at scale, and delivered through the company’s subscription platform. The ability to connect research with field realities is ultimately what transformed new ideas into trusted commercial technologies.
NA: How has your SPE involvement shaped your research direction or opened doors you might not have found otherwise?
US: SPE has been much more than a professional organization for me; it has served as a gateway to the wider energy community. In an industry that is surprisingly small and closely connected, it introduced me to mentors, collaborators, and friends whose advice often influenced my research direction. Serving as a reviewer and technical committee member exposed me to emerging ideas and helped me identify industry problems that still needed better solutions. Leadership and conference roles strengthened my ability to communicate, organize, and bring together people with different perspectives. These relationships also opened doors through collaborations, referrals, and recommendations that I might never have found otherwise. Overall, my involvement with SPE has been deeply fulfilling and rewarding, and it has played a pivotal role in shaping my career.
NA: As ML capabilities, as well as AI in general, grow, how do you think the role of the reservoir/production engineer itself will change?
US: Technical expertise and domain knowledge will always remain the foundation and the most important qualities a reservoir or production engineer brings to the table. What is changing is the set of capabilities built upon that foundation. Data science, programming, and modeling skills are already becoming core requirements in engineering roles, and that trend will only accelerate.
I believe the next step is agent assisted engineering, with production and reservoir modeling advisors acting as intelligent collaborators. Imagine describing a problem through a chat interface and having the agent prepare the input deck, run simulations, compare scenarios, and summarize the results. Glimpses of this future are already appearing in newer releases of leading commercial simulators. Soon, direct connections between simulators and AI systems such as GPT and Claude, through frameworks like MCP, could add a new layer of automation, accessibility, and creative problem solving. However, the quality of the output will still depend heavily on the engineer guiding the agent. Strong domain experts will provide richer context, ask sharper questions, identify weak assumptions, and guide the agent toward more informed results while ensuring its recommendations remain physically and operationally sound. I see this as the natural evolution of the profession, not a threat to it. Every generation of engineers has adapted to the technologies of its time, just as engineers once moved from pen and paper to computers. AI adds speed, scale, and efficiency, but it does not replace engineering judgment or experience. The engineers who embrace it as an amplifier of their expertise will be the ones who shape the profession’s future.
NA: In addition to your corporate role, you collaborate with Birol Dindoruk at Texas A&M University as a volunteer researcher. Could you tell us about this collaboration, the research you pursue, and how your academic work complements your corporate career?
US: I could speak about this collaboration at length, but I will try to keep it brief. Birol Dindoruk is internationally recognized for his work on reservoir-fluid phase behavior, gas-injection improved oil recovery, and CO₂ sequestration. His work has earned some of the profession’s highest distinctions, including SPE Honorary Membership and election to both the National Academy of Engineering and the National Academy of Inventors. His reputation, together with my own background in and interest in fluid thermodynamics, initially drew me to his class. However, it was his depth of knowledge and high expectations that inspired me to continue working with him. When I was unable to take another course with him, I pursued an independent research project on mathematically modeling the effect of dispersed asphaltenes on heavy-oil viscosity.
At the end of that semester, he told me the work had publication potential. I continued developing it, and within two months, our paper was accepted with almost no changes by the as Geoenergy Science and Engineering. That experience became the foundation of a collaboration that has spanned more than 9 years, including 8 years after graduation, during which I have continued working with him as a volunteer research associate. What began as a single independent study has grown into nine journal papers, five conference papers, several freely available engineering tools, and continued collaboration through conference committees and technical reviewing. Our research combines advanced data-driven methods, including deep learning, with traditional petroleum engineering to address complex problems in fluid PVT behavior, enhanced oil recovery, and reservoir modeling and simulation.
My academic and corporate work complement each other naturally. Research allows me to explore fundamental questions, investigate new ideas, and stay current with emerging scientific and technological developments. My industry experience, in turn, keeps those ideas grounded in strong technical fundamentals, domain knowledge, operational realities, and practical value. Dindoruk has been more than a research advisor; he has been a steady mentor, motivator, and source of perspective. He has taught me patience, rigor, curiosity, and the importance of making technical work useful to others. What began as a classroom connection has become one of the most meaningful and enduring professional relationships of my life.
NA: As a TWA Energy Influencer, what is your advice for students and young professionals who are interested working in the energy industry?
US: Build a strong foundation in your core engineering discipline while developing coding, data, and digital skills alongside it. Stay aware of trends across the broader technology market, because many of those advances will eventually reshape the energy industry. Attend conferences, technical sessions, and free lectures from leading professionals, and listen with genuine curiosity. The more knowledge and perspectives you absorb, the better questions you will ask and the more thoughtfully you will approach problems. Build authentic friendships and professional relationships through organizations such as SPE and never be afraid to take initiative before you feel completely ready. Every new skill, conversation, and courageous step can create opportunities you cannot yet see.
