Research direction
I am a PhD researcher in computational mechanics and scientific machine learning, working at the point where numerical methods, high-performance computing, and learning-based models can inform one another. My research focuses on phase-field fracture, differentiable simulation, inverse problems, and the use of machine learning in mechanics. I am registered at City, St George's, University of London, with teaching and research-facing work at Queen Mary University of London.
My aim is to make high-fidelity mechanics more useful in practice: faster to run, easier to inspect, and able to work directly with gradients and data. That direction connects fundamental fracture mechanics with engineering software, aerospace structures, and AI for science.
Research and software
I developed PhAST, a PyTorch-native, GPU-accelerated differentiable physics engine for phase-field fracture. The work brings explicit dynamics, staggered damage solves, and autograd-compatible inverse analysis into one open research codebase. The associated preprint, A matrix-free, differentiable PyTorch solver for phase-field fracture, presents the formulation, benchmarks, and inverse-analysis examples.
I am lead author of Machine learning for computational fracture and damage mechanics: status and perspectives, published in Engineering Fracture Mechanics with collaborators at EPFL, the University of Florida, and City St George's. The review maps how machine learning is being used across fracture and damage mechanics, while distinguishing credible opportunities from open questions in validation, data, and physical consistency.
Engineering background
Before doctoral research, I founded Aeroknacks, where I built aerospace structural-analysis and hand-calculation automation tools. That work translated established design references, including E. F. Bruhn, Michael Niu, Boeing Design Manuals, and ESDU data, into practical Excel-VBA and engineering-software workflows. It covered fastener load transfer, lug strength, plastic bending, buckling, composite laminate calculations, and bolted-joint stress fields.
The open-source BJSFM project grew from this experience. It implements analytical stress-field methods for anisotropic composite plates and reflects the kind of software I enjoy building: technically grounded, transparent, and useful to engineers.
Teaching and communication
I contribute to engineering teaching at Queen Mary University of London, including mechanics, materials, manufacturing, wind-turbine structural design, and Siemens NX CAD activities. I also created and maintain open tutorials in PyTorch, neural operators, physics-informed neural networks, and related fundamentals to support Dr Sathiskumar Anusuya Ponnusami's teaching while making the material useful to a wider audience.
My teaching support was recognised with the 2026 School of Science and Technology Dean's Award for Outstanding Teaching Support. I value clear exposition as part of technical work: an implementation or model is more useful when someone else can understand, reproduce, and question it.
Education and recognition
I hold an MTech in Design and Manufacturing from the National Institute of Technology Silchar, where I graduated with a CGPA of 9.88/10 and received the Best Student Award. I previously completed a B.E. in Mechanical Engineering at GM Institute of Technology, Visvesvaraya Technological University.
My doctoral work is supported by a fully funded City, St George's studentship. I received the 2026 Yeoman and Travelling Scholarship from the Worshipful Company of Tin Plate Workers alias Wire Workers to support conference travel and dissemination, and I am an Associate Fellow of the Higher Education Academy.
Current interests
I am particularly interested in roles and collaborations involving differentiable simulation, scientific machine learning, computational mechanics, AI for engineering, and engineering research software. I work primarily with Python, PyTorch, CUDA-capable GPU workflows, finite elements, optimisation, and reproducible research tooling.