In this project, we wanted to design a recommendation scheme for suggesting relevant events/products to a mobile user based on his/her interests and activities. The recommendation could be provided based on the user's spatio-temporal location or his/her social group behaviour. For this purpose, we have to perform data-mining on the movement patterns of the user, to create his/her profile, and at the same time, identify various common interest groups a user belongs to. I designed an event-driven simulator to generate realistic movement patterns of individual users as well as groups of users, to provide synthetic data for testing mining algorithms.
ACADEMIC- RECORD
Cornell University (2011-Present)
- 2nd Year PhD student, Computer Science
- Cumulative GPA: 4.06/4.0
Indian Institute of Technology (IIT), Kharagpur (2006-2011)
- Dual Degree: B.Tech (Hons.) and M.Tech(Hons), Computer Science and Engineering
- Cumulative GPA: 9.63/10.0 (Ranked 1st in the institute)
B.J.B. Junior College, Bhubaneswar (2004-2006)
- Higher Secondary Certificate (Science), CHSE Orissa
- Aggregate: 92.4% (Ranked 3rd in the state of Orissa)
Stewart School, Bhubaneswar (1998-2004)
- Indian Certificate of Secondary Education, CISCE New Delhi
- Aggregate: 95.7% (Ranked 5th in the state of Orissa)
ACADEMIC PROJECTS
1. Analytics of Group Dynamics of Mobile Phone Users (Sponsor: Xerox Co.)
Masters Thesis Project
2. Fault Tolerant Analysis of Automotive Systems (Sponsor: General Motors)
Bachelors Thesis Project
In this project, we developed a novel approach for performing quality centric analysis of automative systems. The approach uses two fault analysis methods in tandem: a fault-injection and simulation method to evaluate the quality degradation of output signals of the system, under specific testcases and fault-scenarios; and a look-up table based approach, to obtain the quality degradation of the complete system by a series of table lookups, which store the characterized quality degradation behavior of individual components. Using these techniques, I improved the characterization of several individual components of the system.
3. Designing a Bluetooth-Twitter Interface
Design Lab Project
In this project, I designed a Twitter interface based on Bluetooth communication. Each device could follow any number of devices in range. Whenever a user performs a status update, it would be visible on the wall of all his followers. I also computed statistics like updation frequency, inter-contacts duration and transmission latency of my design.
4. Traversing NAT Boxes and Firewalls: A peer-to-peer Requirement
Term Project, Ubiquitous Computing
In this project, I studied several NAT traversal protocols such as STUN, TURN and UDP Hole Punching, and implemented a chat client for communication between machines behind NATs (like Skype). Apart from this, I also analyzed the design challenges in building large scale p2p based Video-on-Demand systems regarding the appropriate replication strategy, piece selection techniques, and transmission strategies.
5. Analysis of Query Logs and Word Co-occurrence Networks
Term Project, Complex Networks
In this project, we analysed the impact of using standard document-analysis techniques to pre-process queries for IR systems. We constructed restricted and unrestricted Word Co-occurrence Networks from the AOL Query Log Corpus and Europarl corpus, and compared the different networks based on their degree distribution, clustering co-efficient and spectral plots.
6. Hybrid Query Execution Model for Relational Database Systems
Term Project, Database Management Systems
In this project, we combined the techniques of data shipping and query shipping to design a hybrid query shipping model for optimized performance. We formulated an execution plan for the set of queries, annotating each with the site of its execution (client or server) so as to minimize the time of execution. To achieve this, we built a cost-effective query optimizer based on the Volcano optimizer and obtained improved query response times.
