Singapore
Spends majority of time on software development which I enjoy tremendously and continue to pursue. Capacity for advanced research and software development at a deep level. Appreciation of problems from a business point of view. Independent and adaptable to various verticals. Believes in achieving higher efficiency in work through mutual mentorship and guidance of others in the team. Specialties: Machine Learning, Computer Vision (2D & 3D), Target Tracking, Probabilistic Robotics, Deep Learning, Graphical Models, OS, C/C++, Python, Multithreading, Docker,
Responsibility consists of software development and implementation of new algorithms and features, Taking point in leading efforts on algorithmic developments and also on the architecture of the video analytic product supporting a large number of camera streams. Using techniques from Computer Vision, Machine Learning and Deep Learning .
Software development on core video analytic product, the development of its algorithms for new features and performance. Taking lead and actively developing software related to video analytics, from research to product. Using techniques from Computer Vision, Machine Learning and Deep Learning .
• Set up a software development team from scratch • Obtained Spring TECS POV Grant. Responsible for technical details. Grant was to develop a commerical software for the detection of anomalies in videos. • Gained industry specific experience on CCTV setup and architectures. • Wore multiple hats, from hiring to project management, system administration and technical direction. • Software team consists of mostly junior developers but prototype was developed in 9 months. • Successful completion of Spring TEC POV • Part of team that raised pre-series A round for startup Vi Dimensions.
• Python bindings for a large commercial C++ library for statistical inference with SWIG. Includes unit testing. • Implementation and extensive experimental evaluation of algorithms based on the Hidden Markov Model for behavior recognition/time series classification. • Data scrubbing and time series filtering with kalman filters as part of a longer data processing pipeline. • Parallelization (via multithreading) of an often used machine learning algorithm within the company. Parallelization resulted in an approximately 6x reduction in execution time on a server with 8 cores. • Design and implementation of probabilistic algorithms, for a toy robot, enabling it to navigate and localize itself in real time. A simple middleware and firmware for the robot was also developed to facilitate synchronization (with the probabilistic algorithm) and communication of data over zigbee. • R&D of a proof of concept for a camera based tracking of parcels in challenging scenarios. Extensive use of the library OpenCV enabled the rapid prototyping, bringing research ideas to proof of concept in 2 weeks. • Implementation of several different algorithms for the stitching of aerial imagery taken from drones.