Perception Machine Learning Engineer
Medical Devices
DEKA Research and Development Corporation

During my tenure, I contributed significantly to various projects involving innovative medical devices:

Infusion Pump

I engineered a sophisticated system incorporating vision detection algorithms, particularly Hough transforms, to analyze RGB images from the pump’s camera. This system accurately tracked drops and streams, precisely calculating liquid flow rates. Additionally, I developed a versatile cost function to facilitate informed decision-making based on diverse detection outputs. To synchronize flow estimation outputs with ground truth data, I employed dynamic time warping (DTW) and Fourier transforms, ensuring accurate assessments.

Insulin Pump

My involvement included designing a personalized differential equation model to compute tailored insulin doses based on continuous glucose monitor readings, considering factors like meal times, types, age, and gender. Currently, I’m exploring advanced techniques such as Reinforcement Learning, Transformer, and GAN on time series data to enhance insulin pump autonomy and reliability, aiming to optimize insulin delivery without relying on user-input meal information. Moreover, I developed a computer vision method to identify weld defects during device manufacturing, employing various texture detection methods and Euclidean distance transform to assess weld thickness and identify defects.

Kidney Transportation Box

I played a pivotal role in developing a segmentation model aimed at monitoring changes in kidney size and color, offering insights into kidney health for both pig and human kidneys. This model provides valuable information for doctors to better understand and assess the condition of the kidneys.

Throughout these experiences, I collaborated closely with interdisciplinary teams, applying cutting-edge technologies and methodologies to enhance the functionality and reliability of these critical medical devices.

More information can be found here DEKA