Jan 2019 – Aug 2020 · 1 yr 7 mos

Feb 2016 – Aug 2020 · 4 yrs 6 mos

Feb 2016 – Jan 2019 · 2 yrs 11 mos

Jun 2014 – Sep 2014 · 3 mos · Redmond, Washington

• Implemented Supervised Semantic Indexing for ordering results produced by conversational systems, and for choosing the best natural language template that can be used as a reply.
• Designed a graphical interactive tool using C# that enables system designers to train the ranking algorithms even without machine learning expertise.

Jun 2011 – Sep 2011 · 3 mos · Santa Clara

• Designed a security algorithm for anonymous handling of data on a cloud without disclosing it to even the server. The task presents difficulties for cryptography algorithms which generally rely on key sharing.
• Employed different machine learning techniques to analyze a workloads signature (resource consumption along several dimensions). Designed an algorithm which predicts resource consumption and hence enables real time scheduling.

Sep 2010 – Dec 2015 · 5 yrs 3 mos

• Computational modeling of human/animal behavior in different domains with the two-fold aim of gaining insights into the working of the brain, and to develop behaviorally inspired machine learning algorithms.
• Projects include modeling active sensing as Bayes-optimal sequential decision making; decentralized, dynamic model for the problem of competitive foraging; and ongoing work on modeling emergence of contextual effects in multi-attribute choices.

Jan 2010 – Jan 2015 · 5 yrs

Jun 2008 – May 2010 · 1 yr 11 mos · Bangalore

• Designed and implemented an algorithm for Time Division Multiplexing over IP that can achieve synchronization across remote sites with very high precision.
• Solved a related problem of playout buffer size using an adaptive playout mechanism so that latency and packet losses can be minimized.
• Designed and implemented algorithms for noise suppression and echo cancellation as part of the project on voice quality enhancement.

May 2007 – Jul 2007 · 2 mos · Clemson, South Carolina

• Proposed a new mobility model as an improvement over the classic Random Way Point model. The new model achieves the desirable uniform distribution on circular geometry and a near- uniform distribution on a toroidal geometry