Projects
Projects
StreetX enables user to represent constraints on multiple arbitrary space regions and time windows using a simple abstract language. StreetX is scalable and is designed to handle large amount of spatio-temporal data from multiple users. Multiple space and time constraints can affect performance of the query and may also result in conflicts. StreetX automatically resolve conflicts and optimizes the query evaluation with access control to improve performance.
Project page (Complete List of Publications and Tools)
In this project, I am working on developing a platform to put water quality data to productive use. We are using IBM Cloudant database to integrate both historical and realtime sensing data at common platform, from which useful information for use cases like drinking, irrigation or bathing can be drawn. We have provided mobile and web based application along with REST API which can provide water analytics to user in realtime.
The mission of the MetroInsight project is to build an end-to-end system for knowledge discovery using highly-dimensional sensor time-series and real-time data streams to support the metropolitan infrastructure through effective analytics, workforce development and policy support. Working with a strategically chosen set of city governments in the Southwest, utilities and companies, we have unique access to specialized metropolitan data providers and National initiatives that have not yet been accessible to pipelines such as those proposed here. The project aims to overcome the data deluge caused by noisy multimodal urban sensory data. It will pursue advances in models and methods to transform multimodal urban data to a lower dimensional population-level data suitable for dynamic processing, real-time monitoring and visualization.
In this project our goal was to develop a salable system which can do real-time analytics of different health conditions. Different health conditions can be regarded as the complex events and thus this concept can be extended to other use cases easily. Large number of users should be able to send the health data in real-time and along with receiving back the feedback and results. Keeping the requirements in mind we used Kafka and Spark to develop our system. Multiple users are like Kafka producers sending data in real-time. Spark streaming is used to process data of different window sizes to analyze the health conditions. In our system we have developed and tested the heart attack risk prediction and stress prediction as our sample complex events. We have simulated and tested our system with multiple health datasets.
In this project, I am working to distribute the dataflow graph of Tensorflow on multiple heterogeneous devices depending on the device capabilities, energy and cost of different computing nodes.
Log Query Interface is an interactive web application that allows users to query the dataset easily and efficiently. With this interface, users no longer need to talk to the database through command line queries, nor to install the MobileInsight client locally to fetch data. Users can simply select/type the query message through our website, and wait for the result. The query response time is hugely improved by the server side spark clusters, which stores the big datasets in a distributed system.
In this project, We used mathematical modelling to numerically solve the partial differential equations applicable to oil and gas industry. We are using advanced version of finite element method called extended finite element method which can model multiple faults/fractures independent of the mesh. We are developing industrial scale large mesh solver using MPI environment which can handle millions of meshes in 3D. The developed system is continuously developed and tested with benchmarks.
This was my B.Tech Thesis project under the guidance of Dr. Vaskar Raychoudhury (Faculy, IIT Roorkee) during my final year of bachelor. A Smart-phone based health monitoring system (HMS) using body sensors is a useful tool which can do personal health check-up and data collection for remote areas which have absence of medical facilities. This system allows easy to use plug and play interface, mobility, round the clock availability and efficiency in all the process of data collection and maintaining data records. Such a system can be a movable smart hospital in the hands of a professional and in the hands of common man it is like his personal health checker which can do his basis diagnosis and can also transfer the health data to a professional for more detailed analysis. we developed a mobile solution for real time health monitoring. The proposed HMS collects sensed data from body using various sensors and micro-controller that is forwarded to a Smart-phone through USB. Smart-phone is used to display information both in the form of numbers and graphical plots to user. All the process of data acquisition is performed in real time. The proposed system was tested and demonstrated for personal usage and for remote health monitoring. The system was used for Disease Outbreak-Period Detection. The results were presented in IEEE Healthcom 2014 conference.
In this project I worked with Dr. Vaskar Raychoudhury in my third year of bachelor. We developed a new and novel system which was featured in Hindustan times on 8th may 2013. We named it CROWD-PAN-360 (CP360) which generates a fully-tagged 360 degree panoramic map of the surroundings of a querying user using crowd-sourced images, audio trails, object tags, and raw location data collected by smart phones in an opportunistic manner. Recent advances in smart-phones and location-aware services necessitate identifying logical locations of users, in terms of their surroundings, instead of raw location coordinates. The objects (logical locations) appearing in the images are identified using manually or automatically generated tags. The system is context-aware and it intelligently associates user location coordinates with several smart-phone contexts, like acceleration and orientation. CP360 can significantly reduce GPS positional errors for even cheap low-end smart-phones and can identify the user surroundings very efficiently. We extensively tested the system in both indoor and outdoor environments of IIT Roorkee campus using Android smart-phones over a dataset of more than 6000 crowd-sourced images of nearly 70 objects(departments, hostels, cafeteria, etc.) and CP360 generates the panoramic map with an average accuracy of 92.2%. The results were accepted in journal, IEEE TPDS 2014.
This was project done as part of IBM National Technical Challenge (NTC 2013). Our work was accepted among the five teams in finals, held on October 19, 2013. In this research problem we worked on finding the top crime prone areas using GPS traces of criminals. find crime suspects among civilians and criminals using their GPS traces. The challenges in the problem were due to mobile phone limited processing capabilities. Due to large volume of data most of the clustering algorithms are very resource intensive and cannot be directly used for mobile devises. We implemented a modified version of K-Means clustering on android and developed parallel implementation which could analyse multiple trajectories of criminals simultaneously. Initially K sample points are selected. K is kept very large so that finally we can overcome effect of selection of distributed K points. The K clusters are grown in size up to the maximum limit. The density parameter of each cluster is calculated. Finally duplicate clusters are removed and clusters are sorted according to densities.