Apr 2013 – Nov 2017 · 4 yrs 7 mos · San Francisco Bay Area

• Built and managed multiple teams of scientists focused on applying AI, Machine Learning & Differential Privacy to fight account takeovers (ATO), fake accounts, financial fraud on the App Store & Apple Pay, search abuse, fake reviews, App Store piracy & iMessage spam. A sampling of projects:
• ► Unsupervised and supervised transfer learning using auto-encoders & deep neural nets for estimating account trust.
• ► Deep Learning using Recurrent Neural Nets for event sequence mining to secure accounts.
• ► Rank-Support Vector Machines & text embedding (NLP) for the App Store review relevance scoring.
• ► Machine Learning with click stream & text data to increase the App Store search relevance.
• Patents:
• Differential privacy for message text content mining (US20180091466A1)
• On-device rate-liming using tailored proof-of-work (US20180089465A1)
• Electronic device passcode recommendation using on-device information (62/738,698)

Feb 2011 – Apr 2013 · 2 yrs 2 mos · San Jose, CA

• ► Mining data for similarity, Collaborative Filtering with missing data, Parallel Machine Learning, Predictive Analytics with text, image & video data
• ► Originally conceived, nurtured and evolved (in collaboration with two team members) a research idea into an IBM Grand Challenge (some other IBM GCs have included Deep Blue, Jeopardy Watson). Over 20 researchers working on the project across three countries.
• ► Worked closely with a major client as we deployed high value analytics solutions tapping into client’s data banks. Provided the technical analytics backbone for the IBM team.

Jan 2010 – Jan 2011 · 1 yr · Cambridge, MA

• ► Invented Radon-Like Features (Connectome Project, MMBIA 2010, with A.V. Reina & H. Pfister) for neuron boundary segmentation in Electron Micrographs. Since then, Radon-Like Features have been successfully used for satellite image analysis, angiogram vessel tree analysis and echocardiogram Doppler image analysis.
• ► Invented Kernel Plurality (ICCV 2011, with A. Banerjee, B. C. Vemuri & H. Pfister), a method for combining multiple classifier outputs or building consensus among voters, in a supervised setting, such that the victory margin for the correct class with respect to all other candidate classes is maximized.

May 2009 – Aug 2009 · 3 mos · Cambridge, MA

• Invented MoRF: Morphable Reflectance Fields (CVPR 2010, with M. Jones & T. Marks), a method for enhancing single image face recognition by augmenting single image galleries with automatically generated new relit images. Additionally, built a micro-controller synchronized multi-camera multi-flash face image acquisition dome for data collection.

May 2008 – Aug 2008 · 3 mos · San Jose, CA

• Medical Image Analysis, Echocardiogram Analysis, HARDI MRI Reconstruction

Aug 2005 – Dec 2009 · 4 yrs 4 mos · Gainesville, Florida Area

• ► Worked in various capacities (Teaching Assistant / Research Assistant / Graduate Assistant) as an Alumni Fellow. Alumni Fellowship is awarded to "Graduate students with outstanding undergraduate preparation, a strong commitment to their field of study and demonstrated potential in research and creative activities" www.uff.ufl.edu/scholarships/
• ► One semester my duties included brewing coffee for the department office. A professor in appreciation of my work once commented "Ah, it's so great to see someone correctly label this coffee COLOMBIAN, I was so sick of seeing the incorrect COLUMBIAN tag on it".

Aug 2004 – Apr 2005 · 8 mos · Gandhinagar, Gujarat, India

• Assisted professor with classroom teaching, grading and research projects

May 2003 – Jul 2003 · 2 mos · Mumbai Area, India

• Studied cellular communication protocols