Mehmet Kerem Turkcan

Greetings! I am currently an associate research scientist in the Center for Smart Streetscapes (CS3) and Civil Engineering & Engineering Mechanics at Columbia University specializing in computer vision and deep learning applications. Before, I was a postdoctoral research scientist in Electrical Engineering. My research focuses on real-world, low-latency deployment of object detection and trajectory prediction models, and retrieval-augmented generation via large language models. To test algorithms, I lead projects on large-scale data collection and annotation, and improve upon the state-of-the-art models to adapt them to the strict performance and quality requirements of real-world use cases. Right now, my focus is specifically on applications to urban streetscapes and robotic surgeries.

Previously, as a member of the Bionet group, I studied the efficient initialization and execution of very-large-scale, real connectome and synaptome driven, single-synapse-level simulations of neural circuits on GPUs for the purpose of understanding realistic neural networks. Using the tools I developed, I worked on the design of computational circuits for understanding the mushroom body and the lateral horn, two brain regions in insects linked with associative and evolutionary memories. I am also one of the main developers of an open-source interactive neuroscientific computing platform called FlyBrainLab (FBL) that enables access to, manipulation of and simulations for fruit fly brain data.