I am Dr. Seshu Kumar Damarla. I build production-oriented machine learning solutions for soft sensing, process monitoring, and control applications using real plant and lab data. I have delivered industrial analytics work with Syncrude, Suncor, Shell Canada, and Nutrien, and I focus on solutions that remain reliable under noise, missing data, collinearity, and process drift.
Integrating machine learning with control and monitoring: robust soft sensors, valve stiction detection, MPC loop analytics, and data-driven diagnostics for large-scale industrial plants.
I work at the intersection of process engineering and applied data science, translating plant data into maintainable ML solutions for monitoring, soft sensing, and decision support.
I have hands-on experience building data-driven models for industrial processes in energy and chemical facilities. My work covers soft sensors, process monitoring, and control analytics, with emphasis on robustness, interpretability, and practical adoption by operations and APC teams.
I enjoy taking raw historian data, vibration signals, spectroscopy measurements, and lab results and turning them into deployable, monitored models that support decision-making (e.g., early detection, quality estimation, and improved control performance).
I am particularly interested in roles where I can work closely with process engineers and data scientists to design practical, maintainable ML/AI solutions for real plants.
Supervised & unsupervised learning, anomaly detection, deep learning for signals, time-series modeling, and model explainability tailored for industrial processes.
Control loop performance monitoring, valve stiction detection and quantification, APC/MPC analysis, and chemometric monitoring for quality and fault diagnosis.
Python, MATLAB/Simulink, SQL, and modern ML libraries, with experience building end-to-end pipelines from data preprocessing to deployment-ready models (experiment tracking, versioning, and lightweight serving).
A sample of projects where I applied ML, statistical modeling, and control to solve practical problems in oil sands and process industries.
Industrial process analytics • robust modeling • monitoring-ready design.
Built a data-driven soft sensor to estimate unmeasured quality variables using historical plant data under practical constraints (noise, missing data, and collinearity). Benchmarked baseline regression/PLS models against deep learning approaches to improve robustness and stability, and designed the solution with online inference, model monitoring, and retraining considerations for deployment.
Soft-sensor modeling for extraction facility centrifuges.
Developed and deployed soft sensors (predictive models) to estimate the ratio of naphtha and bitumen in the feed stream of centrifuges in the extraction facility at Syncrude Canada Ltd., improving monitoring of feed quality and supporting better operational decisions.
Empirical data-driven approach for slurry preparation plant.
Devised an empirical method for detecting sanding in hydro-transport pipelines in the slurry preparation plant at Syncrude Canada Ltd., using process data patterns to indicate abnormal transport and potential plugging risks.
Multivariate statistical monitoring for upgrading facility.
Formulated a method based on multivariate statistical techniques to detect foaming in amine contactors in the upgrading facility, enabling early detection and corrective action to maintain absorber performance.
Quality prediction from spectroscopy data.
Developed inferential models using Near Infrared Spectroscopy data to predict gasoline properties for Suncor Energy Inc., enhancing product quality assessment and enabling faster feedback than lab measurements.
Decision-support system for operator actions.
Led an Advanced Process Control Decision Support project for Shell Canada, building tools to help control engineers interpret operator actions on critical process variables and improve APC utilization.
From DCS screenshots to controller-ready models.
Developed a graphical user interface to extract data from screenshots of a distributed control system and identify process models, which were subsequently used to design PI/PID controllers for Nutrien operations.
A small selection of IP and publications most relevant to applied data science and industrial process analytics. Full list available on Google Scholar.
US Patent US20240376998A1, 2023
Practical methods for detection and quantification of valve stiction using routine control loop data, enabling targeted maintenance and improved control loop performance.
CRC Press, Taylor & Francis, 2018
Monograph on modeling, simulation, and control of fractional order processes, including identification and controller design techniques for complex process dynamics.
CRC Press, Taylor & Francis, 2017
Co-authored book on chemometric techniques for product quality assessment, process monitoring, and fault detection across industrial applications.
Control Engineering Practice, 2024
Survey and practical guidance on applying machine learning to industrial sensing and control, with emphasis on deployment, data challenges, and real-world constraints.
Can. J. Chem. Eng., 2023
Image-based method using Markov Transition Field and deep convolutional neural networks to detect control valve stiction, leveraging time-series-to-image encoding.
Ind. Eng. Chem. Res., 2023
Developed statistical test-based methods that provide practical and implementable approaches for detecting and quantifying stiction in industrial control valves.
IEEE Transactions on Instrumentation & Measurement, 2023
Proposed a ConvLSTM and self-attention enhanced canonical correlation analysis framework for multi-output soft sensor modeling in complex industrial processes.
Editorial and peer-review service for journals and conferences in industrial data science, process control, chemical engineering, and instrumentation.
I contribute to the scientific community through editorial responsibilities for high-impact journals in industrial informatics and process systems engineering.
Regular reviewer for journals and conferences covering process control, soft sensors, chemometrics, industrial AI, and condition monitoring.
I am open to roles in industrial data science, advanced process control, soft-sensor development, and ML/AI for process industries. If your team works with historian data, control loops, or plant-wide optimization, I would be happy to discuss how I can contribute.
When contacting me, feel free to include a brief description of your process, data sources, and current challenges so I can propose concrete ideas from the outset.