I am a Computer Science graduate student at the University of Illinois, Urbana-Champaign. I have completed my Bachelor of Technology in Computer Science and Engineering from the Indian Institute of Technology, Bombay. I love to tinker around with Computer Systems and excited about the Internet of Things. I am a curious coder and a skilled vocalist.
GPA: 4.00/4
Overall GPA: 9.00/10
Proceedings of the 11th Workshop on Wireless Network Testbeds, Experimental evaluation and Characterization ·
Emerging technologies and applications in Internet of Things are generating large amount of data at the network edge. Machine learning models are often built from the collected data, to enable the detection, classification, and prediction of future events. Due to bandwidth, storage, and privacy concerns, it is often impractical to send all the data to a centralized location like Cloud. Hence it becomes imperative to learn the machine learning models on the Edge in a distributed fashion. I pursued a simulation-based approach under the guidance of Prof. Umesh Bellur to simulate the learning of such models on the Edge.
P4 is the buzzword in the world of networking. It literally stands for Programming Protocol-independent Packet Processors. The routers at your home are hardcoded to process a specific set of network protocols. But with P4, one can PROGRAM network switches (technically, the forwarding plane) to adhere to different protocols. You can read more about P4 here. I worked with Rinku Shah, Akash Trehan, Prof. Purushottam Kulkarni and Prof. Mythili Vutukuru in building a Stateful Load Balancer. Our project started with the intention to compare the performance of the load balancer in a distributed v/s a centralized controller setup. We used P4-14, bmv2 software switch and Mininet for our experimentation. As P4 is a fairly new programming language, we had a steep learning curve - debugging via pcap traces was fun though! As we were developing our application, we realized that a lot of code was getting repeated. We also realized that there is a need to build a library for doing general purpose programming chores, just like C++ or Python has. Thus began our attempt to embellish P4 for our purposes and motivated us to build a preprocessor, pcube, which provides primitives for aiding code development in P4. P4 is going to be THE NEXT BIG THING as programmable chips could play a vital role in 5G networks and the development of the Internet of Things.
Has it ever happened to you that you receive a voice call and the network connection is not quite good? A very common problem faced almost everywhere in India - poor network connectivity to make/receive a voice call from every point inside a building or home or campus. The most common practice to overcome this is to just go to your balcony to receive your calls. Instead of going to your balcony, we can leverage the existing Wifi routers to relay voice calls over the local network. I worked on this project with Aman Jindal, under the guidance of Professor Bhaskaran Raman, where we tried building an Android application which will relay incoming voice calls to trusted devices over a single router. Forming a secure group in a local network is very important as you don't want your calls being tapped by other devices. Hence, I designed and implemented a distributed Device Discovery protocol which uses a ping-like mechanism to alert all other devices in the local network available to be paired. I also designed and implemented a Device Pairing protocol for secure group formation in the local network. Unfortunately, our progress is hindered because Google has completely blocked call recording in Android 9 Pie [source]
Heard about Amazon Go? They are using Computer Vision and Machine Learning to track their customers! Why not use GPS though? Modern GPS systems have an error margin of 3 metres, but these fail miserably while localizing devices indoors. One of the ways to localize indoors is to use Bluetooth Low Energy (BLE) beacons. But if one deploys a layout of beacons in a retail store, the bluetooth signals are often attenuated, e.g. Multipath propagation. To investigate this, I carried out indoor localization experiments in the Grainger library (because the stacks in a library simulate the attenuation effect) using a layout of Bluetooth Low Energy beacons for different power and frequency of packet transmission. I collected and parsed stationary and dynamic data using a packet sniffer which was used in building a packet reception model. I also built an Android application to capture BLE advertisement packets as using this packet sniffer was very cumbersome. But I was not able to implement channel-specific packet filtering due to the hardware limitations of our smartphones, as pointed out here by Simon Judge. I would like to thank Prof. Hari Sundaram for providing me with the opportunity and Subham De for helping me out in this project.
Conducting lab sessions and addressing the queries of students regarding lab assignments. Designing course for the Honors class and mentoring students for a semester-long project.
Assisted Prof. Ganesh Ramakrishnan in setting lab assignments and cleared difficulties faced by freshman students in the lab
Assisted Prof. Supratim Biswas in setting programming assignments and prepared automated assignment grading scripts
Assisted Prof. Bhaskaran Raman in grading exams and lab assignments, and handling course logistics for IIT Dharwad
Assisted Prof. Umesh Bellur in setting lab assignments and cleared difficulties faced by freshman students in the lab
Click here to see my Curriculum Vitae