I did my PhD from the School of Informatics, University of Edinburgh under Dr. Rik Sarkar in "Machine Learning and Privacy Preserving Algorithms for Spatial and Temporal Sensing" [Thesis]. I was a Research Associate in the Department of Computing, Imperial College London under Dr. Thomas Heinis. My research interest is to design analysis and privacy-preserving algorithms using tools from topology, geometry, and statistics to infer abstract knowledge from massive and noisy sensing data.
Periodicity detection is challenging in a noisy strem where periodicity and phase change and it has false positive events. Here, we propose a Particle filter based perioricity detection method for such streams.
An efficient framework to map trajectories to Euclidean vectors called signatures using differential topology. Analytics on signatures are more accurate and efficient than machine learning on raw data; Especially, motion prediction at large scale (e.g., 500 meters). Signatures preserve enough information to reconstruct non-self-intersecting trajectories.
Proposes private and efficient data structure to answer range queries along paths in a planar graph. The data structure pre-computes range counts in canonical paths (a small subset of the possible paths) using notions from planar separation theorem and random sampling. The query satisfies differential privacy and beates state of the art local differential privacy in utility. The structure supports a restricted class of paths based on shortest paths and extends to 2D range queries using differential forms.
Various intervention methods have been introduced worldwide to slow down the spread of the SARS-CoV-2 virus, by limiting human mobility in different ways. While large scale lockdown strategies are effective in reducing the spread rate, they come at a cost of significantly limited societal functions. We show that natural human mobility has high diversity and heterogeneity such that a small group of individuals and gathering venues play an important role in the spread of the disease. We discover that interventions that focus on protecting the most active individuals and most popular venues can significantly reduce the peak infection rate and the total number of infected people while retaining high levels of social activity overall. This trend is observed universally in multi-agent simulations using three mobility data sets of different scales, resolutions, and modalities (check-ins at seven different cities, WiFi connection events at a university, and GPS traces of electric bikes), and suggests that strategies that exploit the network effect in human mobility provide a better balance between disease control and normal social activities.
Supports efficient and parallel range queries by partitioning large spatial datasets into computing clusters. Partitioning ensures that the clusters are load balanced in in terms of data and queries. It uses random sampling and learning methods for partitioning.
Using Bluetooth connection automatically detects when a user leaves her workstation and transfers active applications contexts to phone, for example URL and timing of a Youtube video being watched.
An intermediator router with all supported communication hardware and a software layer to convert between protocols. The router handles the control messages for the concerned protocols and encapsulates the data as the protocols require.
Worked on Bluetooth and Bluetooth Low Energy communication solution for Android and Wearable embedded devices.
Design and develop BLE software stack as a part of Intel’s first BLE chip.
Develop Bluetooth stack and hardware specific drivers for Android phones with Intel Bluetooth chip.
Intel Bluetooth drivers in Linux kernel, advanced BLE data link layer in Android open source (available in all Android devices since Android 7).
Develop automation framework to test Intel communication hardware.
Analysis and Privacy Preserving Algorithms of Spatiotemporal sensing
CPI: 9.49 / 10
Master's thesis: Determination of Safe regions in Optimized Dynamically Updatable Programs
CPI: 9.40 / 10
Ranked 9st out of four hundred thousand students in the state in Higher Secondary (10+2) examination, 2005
Awarded Best Student Project of the institute by TCS Ltd. for B.Tech. final year research thesis (1 yr.), 2009
Scored 10/10 in the M.Tech. final year research thesis (1 yr.), 2011
Received Intel India divisional recognition award (awarded quarterly across Intel India) for conceptualizing Intel’s first smart watch, 2014
Demonstrated feasibility of seamless context transfer in Intel India Innovation Summit, 2013
First to demonstrate Object Transfer Service (OTS) in Android platform across all industries, 2014
Represented Intel Corp. in Bluetooth SIG Interoperability test event as a technology expert, 2015
Received both Principal’s Career Development Scholarship and The Global Research Scholarship (£31.5k / year for three years), Sep ’15 (Only student to receive both in that year from School of Informatics)