Conferences
1. Co-authored a manuscript titled Exploring task placement for edge-to-cloud applications using emulation in 2021 IEEE 5th International Conference on Fog and Edge Computing (ICFEC 2021).
2. Co-authored a manuscript titled Pilot-Edge: Distributed Resource Management Along the Edge-to-Cloud Continuum in 2021 3rd Workshop on Parallel AI and Systems for the Edge (PAISE 2021).
3. Co-authored a manuscript titled Real Time Measurement of Deformability Index for Electro-Mechanodiagnostics accepted as oral presentation and finalist for best student paper award in IEEE International Conference on Nano/Micro Engineered and Molecular Systems (NEMS), Singapore, 2018
4. Presented a poster titled Modeling and Simulation of Microfluidic Viscometer with Integrated Micro-cantilever Sensors for Blood Parameter Monitoring at Institute of Smart Structures and Systems International conference at Indian Institute of Science (IISc) Bangalore conducted byInstitute of Smart Structures and Systems (ISSS).
5. Participated in National Symposium on Human Diseases at BITS-Pilani Hyderabad Campus, 2014.
Research Work
As a member of RADICAL Lab at Rutgers University, I am developing an Emulator as a design and implementation decision support system for distributed data processing and learning for sensors to predict sugar, alcohol, and, Tannins content in red and white wine. I am also part of the team that is working on an unified resource management tool (pilot-abstraction) to manage streaming frameworks for processing and message brokering on High-Performance Computing (HPC) environments.
Developed a software implementation of 4G LTE communication network on ORBIT testbed at WINLAB to emulate the generation of mobile application data. The amount of resources needed for relaying the IP packets for different applications differ in number. If the source is not classified in advance, some minimum amount of resources have to be allocated irrespective of the actual amount required. Classifying application data ahead of time makes this allocation efficient. I analyzed the data generated to select features for classifying user mobile application data into data generated by applications such as Skype, YouTube, Gmail, and Web browsing using k-nearest neighbors, and neural networks. For further information, visit https://deeplearningmethods.wixsite.com/classification.
Developed a DogSheepBot for Introduction to Artificial Intelligence under course advisor Prof. Wes Cowan, Dept. of Computer Engineering, Rutgers. Implemented Markov Decision Process to navigate two Dog bots shepherding a sheep in a 2-D environment. Optimum and Non-Optimal starting positions were generated by calculating the maximum utility value in a state using the value iteration algorithm. Another project, Image colorization, required us to color a Grayscale image into RGB (Red Green Blue) scale image. The problem first required us to extract the most dominant colors (clusters) in the image using K-Means Clustering. This led to a reduction in the input data space. Then, using Artificial Neural Networks (Backpropagation Algorithm) with one hidden layer, we modeled the system to map between Grayscale pixel to RGB pixel.
Developed Maze-Runner (path-finder), Minesweeper game and Probabilistic Hunting project for course Introduction to Artificial Intelligence under course advisor Prof. Wes Cowan, Dept. of Computer Engineering, Rutgers. For Maze-Runner we implemented A* algorithms using Manhattan and Euclidean distance heuristics, and Bidirectional BFS to find a path in 2-D maze. Minesweeper application involved the generating and updating inference at multiple time stamps and implementing it as a Constraint Satisfaction Problem. In Probabilistic Hunting project, implemented Bayesian Inference to predict the position of a target in an environment by creating and iteratively updating a knowledge base using Bayes theorem.
Designed a Flight Management System for course Data Structures and Algorithms to find the shortest path between two locations using graph-based algorithms and compared Breadth First Search, Depth First Search, Dijkstra’s, Greedy Best First Search, and A*. I also quantitatively established that given an optimum heuristic function, A* search outperforms both Dijkstra’s and Greedy Best First search for directed and weighted graphs.
Designed a web application INVEST-GENIE to predict the future stock prices using machine learning algorithms such as Bayesian Regression, Support Vector Regression and Artificial Neural Networks. The prediction was based on historical and real-time data that was retrieved from the Alpha Vantage API, stored in a MySQL database hosted in the local server. The web application was designed using Flask web application framework. The front end for the web application was based on HTML and bootstrap templates. We further implemented a RESTful API which allowed users to access data from our application. Historical and predicted future stocks were utilised in technical indicators like EMA and volatility ratio to act as a recommendation system for customers to buy/sell/hold a stock in the market.
Developed a stock exchange platform, implemented using double linked list, that allows users to trade stocks, construct a dynamic portfolio, and maintain a separate bank account for cash transactions under course advisor Prof. Shiyu Zhou, Dept. of Electrical and Computer Engineering. As part of my project, I built an interface between C++ and MATLAB to plot the variations in the total portfolio value with time. I further integrated the environment with bridge design pattern, separating the implementation of data member Cash Balance, and associated methods by using an abstract class Account . It was also required to maintain the double linked list sorted at all times and I implemented it with the help of a strategy design pattern, prompting user to select one from the two available sorting techniques. Finally, I designed a graphical user interface for the platform using the Microsoft Foundation Library (MFC).
Developed a Web Application to Perform Comparative Analysis of Query Performance between Amazon RDS MYSQL and Redshift under course advisor Prof. Saed Sayad, Dept. of Computer Science. We constructed entity relationship diagrams (ERDs) and relational schemas using database modeling tool ERDPlus to understand the Instacart dataset extracted from Kaggle and uploaded the data to Amazon S3. Our application, Query Analyzer, performed quantitative analysis of the time elapsed to run identical SQL queries on Amazon RDS MySQL and Redshift database. Finally, we demonstrated a query response time advantage of 3X for Redshift over RDS MySQL database through Query Analyzer. RDS is a relational database. It reads data row wise, traversing through all the fields from left to right to search for the required field value. Redshift, on the other hand, is a columnar Database and reads the only those columns that are queried, from top to bottom. But when writing data, RDS outperforms Redshift since adding entire rows is faster.
Worked under Prof Prosenjit Sen, Centre for Nano Science and Engineering, Indian Institute of Science, Bangalore to design a lab-on-a-chip platform for mechanical characterization of red blood cells (RBCs) for a possible mechanical detection of Diabetes Mellitus.
I also worked on a, Electro Wetting on Dielectrophoresis (EWOD) based lab-on-a-chip DNA amplification device, collaborative project between Government of India and Indian Institute of Sciences, Bangalore.
The aim of my project under Prof. Sanket Goel in the Department of Electrical and Electronics Engineering at BITS- Pilani Hyderabad Campus was to design a Microfluidic Viscometer based on Integrated Micro Cantilever to to study the effect of Human blood of varying viscosities on the displacement of piezoresistive Micro cantilever . In this project, I programmatically ran numerical simulations using COMSOL to demonstrate that an increase in % hematocrit level leads to a linear increase in the displacement-stress response of the microcantilever, thus calibrating my device for diagnosis of cardiovascular diseases. Graphical representation was used to relate important parameters such as viscosity of blood, % hematocrit level, stress experienced by the cantilever and its deflection under stress applied. Presented a poster titled Modeling and Simulation of Microfluidic Viscometer with Integrated Micro-cantilever Sensors for Blood Parameter Monitoring at Institute of Smart Structures and Systems International conference conducted at Indian Institute of Science (IISc) Bangalore.
Worked as a research assistant and developed an automated and robust miniaturized platform to conduct Polymerase Chain Reaction (PCR) under Professor Sanket Goel. The lab-on-a chip device will reduced the amplification time of DNA samples and enabled better temperature control than amplification in a conventional laboratory setup.
Worked as research assistant under Dr Debashree Bandyopadhyay, Dept. of Biological Sciences. Our aim was to study the effect of protein micro environment on Tyrosine and its derivatives. Tyrosine is a non-essential amino acid that is the major precursor to Tyrosine hydroxylase (TH), which is the first rate-limiting step in the synthesis of Nor-epinephrine (NE). Individuals with PD have decreased activity of the enzyme TH. We theoretically reviewed all the proteins interacting with TH in Parkinson's disease condition and characterize Tyrosine and its interaction with its micro-environment. This helped us in development of novel techniques for preventing and curing the nuerodegenerative disease.
As a part of Smart Campus project at BITS Pilani Hyderabad Campus, I am developing a mobile platform using SDK 2.2.2 and geo-fencing technology. The vision of Smart Campus is to design a system of interrelated computing devices, cellular phones and people to enable a sustainable and environment friendly campus. My platform facilitates efficient and judicious use of electricity on campus.
Worked as a summer intern from May to July 2015 at the Department of Electrical & Electronics Engineering in the Corporate Social Responsibility arm of the GMR Group on implementing a training program for installation and functioning of CCTV cameras and inverters, which provided me with technical as well as non-profit work experience.