Technology strategy and de-risking for Medtronic's Hugo RAS platform.https://www.medtronic.com/covidien/en-us/robotic-assisted-surgery/hugo-ras-system.html

Working on the Medtronic Hugo Robotic Assisted Surgery system.https://www.medtronic.com/covidien/en-us/robotic-assisted-surgery/hugo-ras-system.html

* Led the Machine Learning Algorithm Development for certified Checkpoint CT product, ClearScan* Led the Machine Learning Algorithm Development for false alarm reduction of certified Checked-Baggage CT product eXaminer* Deep Learning and Classical Machine Learning Algorithm development for certified Checkpoint CT product, ClearScan and certified Checked-Baggage scanner product, MV3D

* Lead the Machine Learning Algorithm Development for certified ClearScan Checkpoint CT product* Deep Learning and Classical Machine Learning Algorithm development for certified ClearScan Checkpoint CT product and certified MV3D Checked-Baggage scanner product* Active DoD Secret Security Clearance.

* Algorithm Development Team Lead for ClearScan Checkpoint CT product achieving industry-first European certification (ECAC C3).* Deep Learning and Classical Machine Learning Algorithm development for certified ClearScan Checkpoint CT product and certified MV3D Checked-Baggage scanner product* Active DoD Secret Security Clearance.

* Threat Detection Algorithms for X-Ray, Skiew-View, and Computed Tomography (CT) data in aviation security space.* Implemented Threat Image Projection (TIP) end-to-end pipeline for carry-on and checked-baggage product lines by inserting 3D scans of threats in clean passenger scanned bags. This capability is used to train screeners and measure their alertness.* Lead algorithm detection effort for implementation of contraband detection from X-ray baggage scans. Performed successful feasibility study and in phase 2, delivered the final algorithms that exceeded customer expectations in detection, false alarm, and throughput rates. Performed large-scale algorithm optimization and Machine Learning approaches distributed on in-house cluster.* Active DoD Secret Security Clearance.

I worked on developing real-time vision algorithms in C++ on Linux platform for robotic application using invariant features, color-based and contour-based representations. I developed object detection and localization algorithms using machine learning for training discriminative classifier, line finding and rectangular region detection. I implemented multiple algorithms for automatic and semi-automatic image segmentation. I implemented algorithm for constrained OCR in noisy and cluttered images.

Working on developing computer vision algorithms, for retail loss prevention, that monitor employee at register for ‘sweet-hearting’ and theft using videos from overhead camerasContributing to algorithms that robustly model employee-object interaction using multiple image feature/region based detection algorithmsImplemented a fast algorithm for cart detectionContributed to the development of client and server side modules to initialize and maintain systems remotely by using Remote Procedure Call (RPC) architecture and database programming

Invented a new method of eyelid/eyelash detection and pixel-wise occlusion mask generation in iris images using log-Gabor filter banks. Implemented real-time algorithms for human iris detection from near infrared image sequences to reject non-eye images/regions and detect human irises within 20 milliseconds per SVGA size imagesWorked on long-range multi-biometric system (Eagle-Eyes) project as software team lead from concept development through proof of concept and demonstrator systemsHelped with managing the Eagle-Eyes project for work break-down, resource assignment, risk analysis and cost structure Worked as technical team lead and project manager to ensure successful completion of initial prototypes, installation and maintenance at client sites. Conducted bi-weekly Scrum meetings during tight project deadlines to keep project on-scheduleProvided top-level management with strategy formulation feedbackLed the system installation/customer training and support for two DoD locationsDeveloped, implemented, customized and contributed to the following computer vision technologies for Eagle-Eyes:Statistical background modeling techniques for real-time human detectionMulti-camera auto-calibration algorithm for multi-camera wide-area scene monitoringCollaborative object tracking across multiple cameras at different resolutionsReal-time frontal face detection & tracking to locate humans of interest in the capture volume tolerant to pose, illumination and appearance changesStable eye detection and tracking algorithms to point iris camera at human eyes for iris acquisition Integration of several third-party face recognition SDK's for face-iris multi-biometric fusionRe-designing of code base on real-time embedded systems principles to use State-Machine architecture for reduced latency and highly scalable system design using publish-subscribe design pattern

Developed a C++-based cross platform application to simulate trajectories of objects in complex activity scenarios for 3-D renderingImplemented a Performance Evaluation Platform for comparing results of object trackers against the given Ground Truth; Implemented and evaluated particle filter, meanshift, ensemble and covariance trackersWorked on Spatiotemporal Video Segmentation in MPEG-compressed video bitstreams using Mean-Shift AnalysisWorked on collaborative object tracking in EPTZ (Electronic Pan-Tilt-Zoom) cameras for wide area surveillance of multiple targetsOff-site

Obtained PhD with dissertation in Content-based video indexing and retrieval using object motion. Used activity representation based on Principal Component Analysis (PCA) and Independent Component Analysis). For efficient motion search, used Spectral Clustering and String Matching techniques. For stochastic modeling, implemented Gaussian Mixture Models (GMM) and Hidden Markov Models. Also addressed the problem of view-invariant trajectory representation. Finally, proposed a method of multiple object activity representation and classification using Tensor Decomposition methods.Worked on Motion pattern Indexing & Retrieval from Video DatabasesDeveloped an object trajectory based video retrieval module Motion-Search to model, search and classify objects based on their motion patternsDeveloped algorithms for Multi-agent Activity Classification using Tensor Decomposition methods as well as Stochastic Modeling for Video Surveillance applicationsContributed to a successful US$0.5M funding proposals based on PhD thesis work resulting in establishment of 'Motion-Search' group at Multimedia Systems Lab, UIC

Worked as summer intern with Data and Sensor group. Researched on video skim generation using unsupervised clustering. Implemented a Windows Media-type application for enhanced browsing of video content based on video summaries for Personal Video Recorders (System and GUI part, implemented on PC for simulation purposes).Developed a 'Video Summarization Tool' in Visual C++ using DirectX and OpenCV for enhanced browsing of video contentDeveloped a Talk Importance Demo to adaptively change playback rate based on Audio-Visual importanceDeveloped a Video visualization system that plays video and updates scatter plots in synchronization for clustering visualizationImplemented video clustering based on Viola-Jones Face Detector results using Spectral Clustering and BiClustering

Worked on implementation aspects of VisionMasterTM , a structured light pattern based non-contact optical system for 3-D solder-paste height measurement used in SMT (Surface Mount Technology) printed circuit boardsHeaded the early research on Automatic Optical Inspection (AOI) system to mark faulty PCBs based on machine vision using robust image registration and alignment.Optimized key image processing modules in the product VisionMasterTM using MMX instruction setBuilt embedded software to interface a USB device using EZ-USB microcontrollers

Developed Visual C++ project for real-time visual extraction of student's ID from answer sheet using Intel's Image Processing Library