Ken Cheng

Throughout childhood, Ken Cheng loved learning about real-world problems that were solved with ingenious engineering solutions. That’s why as a high school student, he valued working on projects based on this idea.

In grade 10, Ken’s inspiration to work on his first-ever research project stemmed from his love of motorsports, specifically Formula 1. Wanting to improve the slow R&D process for testing car parts that involve Computational Fluid Dynamics, he developed a novel testing technique that amalgamated CFD and a backpropagation neural network model that can predict future simulations in a minute fraction of the time it takes to compute all those simulations. This simple, yet effective, project helped him win a Gold Medal at the Canada-Wide Science Fair and a 4th Place in the North American S.-T. Yau High School Science Award. He hopes that this technique can be further developed to improve aerodynamics and consequently, mileage, on all cars.

The following year, Ken's research started with congestion in Toronto. Guided by Dr. Ganesh Mani at CMU, he developed a novel spatiotemporal deep-learning model to predict turning ratios at intersections in a road network. The success behind the model comes from the unique benefits of both components of its base architecture, the Graph Convolutional Network and the Informer architecture, a variant of the widely popular Transformer architecture. This project was selected to be presented at and published in the conference proceedings of the renowned IEEE Intelligent Transportation System Conference. Ken also co-chaired a session at this conference, leading academically rigorous and curiosity-driven conversations about smart transportation for a safer and more efficient tomorrow.

In the summer before his senior year, Ken got to work in Dr. Yu Yang’s lab in the Center for Applied Optimization at the University of Florida through the Student Science Training Program, in which he won the Best Paper Award. Working with HiPerGator, one of the world’s most powerful supercomputers, he was able to find important patterns in the Dual Solutions that were generated via the Column Generation technique for solving the Vehicle Routing Problem. These patterns helped Dr. Yang’s lab design a better Column Generation VRP solver using Machine Learning, with the overall goal of improving the logistics process behind delivering packages in a city.

As an extension of his values in helping others, Ken spent a lot of time in high school mentoring and organizing workshops for over 300+ students to succeed in high school case competitions like FBLA and DECA.

At Columbia, Ken will be pursuing a Computer Science major with a minor in Statistics or Economics, setting him up to create results in research projects and internships that can shatter limits to what we think we can accomplish, helping create a world of possibilities for future generations of scientists.