Indu Kant Deo

I am a Senior Engineer, Machine Learning at Corning Inc., where I work on distributed LLM inference and training performance – building simulators and models for latency, time-to-first-token (TTFT), time-per-output-token (TPOT), and network-aware AI infrastructure. I completed my Ph.D. in Mechanical Engineering at the Computational Multiphysics Laboratory, The University of British Columbia Vancouver, under the supervision of Prof. Rajeev Jaiman, with a dissertation on physics-guided machine learning for dynamical systems, applied to fluid flow and ocean acoustics. My work sits at the intersection of deep learning, physics-based modeling, scientific computing, and high-performance ML systems.

In recent years, I’ve had the privilege of collaborating on projects that combine scientific computing methods with advanced deep learning principles, including:

1. Building a convolutional autoencoder and sequence-to-sequence attention network for long-time-horizon wave propagation predictions, published in Physics of Fluids.
2. Developing a space-time reduced order model that combines 3D convolution kernels to efficiently predict fluid flow dynamics, published in Physics of Fluids and selected as Editor’s pick.
3. Introducing a novel finite element-inspired hyper graph neural network, which enables structured, high-fidelity representation of physical domains, enhancing model robustness and predictive accuracy for complex systems, published in Journal of Computational Physics.
4. Developing a conditional convolutional neural network for real-time transmission loss prediction in varying oceanic environments, with a continual learning approach to account for changing bathymetry.
5. Applying parameterized reduced-order modeling techniques to predict distortion in metal 3D printing, which led to a publication in the NeurIPS Machine Learning for Physical Sciences Workshop.
6. Building an optimization framework which reduces the impact of shipping noise on marine mammals, published in Ocean Engineering journal.

Outside the lab, I’m passionate about the outdoors. Whether hiking up scenic trails, climbing rock faces, skiing down snowy slopes, or swimming in open waters, I find balance and inspiration in nature. These activities allow me to approach my work with a fresh perspective, fostering a problem-solving mindset that’s essential in research.

I’m particularly interested in scientific machine learning, surrogate modeling, digital twins, uncertainty quantification, physics-guided AI, and ML infrastructure/LLM systems – work where machine learning is used to accelerate or improve complex physical or computational systems.