Dhruv Dogra

- Preprocessed Amazon Books dataset as per requirement.
- Implemented an item-item based recommendation algorithm. Evaluated Mean Absolute Error and Root Mean Square Deviation.
- Created a top-10 recommendation list for each user and calculated measures like Precision, Recall, F-measure and Normalized Discounted Cumulative Gain.

Social network to share food recipes built with Rails, Javascript and Bootstrap. Project has been deployed on heroku.

- Enabled login authentication using BCrypt encryption
- Implemented image upload using PaperClip
- Added ability to tag recipes and post comments

- Built a database containing news articles on cryptocurrencies from 200 sources.
- Analyzed and evaluated each news domain based on popularity, total article contribution and content created.
- Evaluated top-100 cryptocurrencies based on three factors namely risk, returns and news coverage.

Technology used: Django, Python, MySQL, HTML/CSS and Javascript.

- Implemented three parallel versions of single source shortest path(sssp) on GPUs using CUDA.
- The first version is a naive bellman ford implementation. Second and third were efficient version where the working set is filtered and reduced in size.

- Developed a web application to predict stock market trends using machine learning algorithms like SVM, Artificial Neural Networks and Bayesian Curve Fitting.
- User could login and manage his/her stock portfolio.
- Bootstrap and Jquery used to create front-end.
- Django framework used to create the entire model.

- Implemented three parallel versions of sparse matrix vector multiplication (spmv) on GPUs using CUDA. C/C++ used to interact with programming interface.
- The first version was a Simple Atomic Based SPMV that used three-step operation. In second version Segment Scan approach, which is based on parallel prefix sum, was used to implement SPMV. Third was a combination of the previous two approaches.

- A web based implementation of neural network for handwritten digit recognition.
- MNIST dataset was used to train the network.
- Python was used to get training parameters
- Web technologies (JS, HTML and CSS) were utilized to create the user interface.

- Setup buffer pool manager and page eviction mechanism (LRU and MRU).
- Implemented Simple-nested loop, Page-nested loop and Sort-Merge joins.
- Used JAVA for coding purposes.

- Developed a system that simulated real-world like terrain containing start and end points.
- Implemented informed and uninformed searching algorithms: : Uniform-cost search, A* and weighted A*(with different heuristic), sequential A*, integrated A*.
- Evaluated and compared the performance of different heuristic functions for A* algorithms.

The purpose of creating the system is improving the customer services by providing a convenience at restaurant table. The objectives are listed below: 

- To construct an interactive multi-touch table for order and other interaction with user which is connected to the server via Wi-Fi. 

- To develop a bot that is able to serve orders from the main counter to the customers’ tables.
- To design a smart prepaid billing system using RFID technology.