SeeScan

Software Engineer (March 2015 - January 2020)

   

● Initiated the first data science team from the ground up and built a linear regression model to estimate battery usage by the various components of the device which helped to increase battery life by 12% in a single battery system and by 18% in a dual battery system.

● Analyzed user logs to understand frequently used features and customer demands. Performed exploratory data analysis and visualized usage patterns to identify the highest captured media type, average video duration, factors that lead to media corruption or sync failure.

● Used Qt to create the user interface for the product and conducted A/B testing to choose the best UI design and flow which significantly reduce the bounce rate by 30%.

● Implemented a failure prediction model and used a sequence mining technique for rules to estimate machine failure on the field using sensor data, which reduced warranty cost by $300k. Detected temperature and fan speed correlation and its effect on the processor clock rate.

● Built a recommendation system to show recommendations for new camera head and reel based on the user’s routine usage of reel length and camera sensor readings which increased sell of reels and camera by 27%.

● Led the transition from non-internet-based device to Wi-Fi enabled device based on data from user logs and multiple sensors.

● Led the transition from non-internet-based device to Wi-Fi enabled device based on data from user logs and multiple sensors.

●  Linux kernel development for Yocto and i.mx6-freescale based pipeline inspection tool. Developed device drivers to control the camera, dual battery system, fan control, and heat management. 


Technologies: Python, Scikit-learn, Pandas, NumPy, SQL, C++, C, Qt, Linux