Atanu Ghosh

Implemented an asynchronous multithreaded producer-consumer priority queuing mechanism (in the form of a system call) which queues jobs by producers in a shared data structure from where consumers can fetch and process them asynchronously. Callbacks from kernel to user space was implemented using netlink sockets. The mechanism supports encryption/decryption, compression/decompression, checksum and concatenation.

Developed an Anti-Malware File System (under Prof. Erez Zadok) on top of Wrapfs File system to intercept read, write, lookup and other systems calls and check against a known malicious pattern database. A user level IOCTL program was written to read and update the pattern database.

This project involved implementation of a new system call in Linux kernel, to support encryption and decryption of a file. Crypto API and OpenSSL library were used. All checks and validations were done at the system level.

This project involved development of a multi-threaded client-server model with a persistent data storage in the form of a shared linked list. The input was basic CRUD (Create, Read, Update, and Delete) queries and the data was stored in simple files. It was tested with over 50000 clients firing random queries at parallel. This project was awarded the ‘Best Project’ award, in the boot camp organized by Oracle in July 2013.

- Implemented a Decision Tree classifier for predicting credit worthiness of Credit Card applicants based on training data. Over fitting and pruning of the tree was done to handle noise in the input data. Missing data in the query could be predicted from the decision tree built.
- Implemented an automated Sudoku Solver using Min Conflicts algorithm and Backtracking with heuristics which include Minimum remaining values (MRV), Forward Checking and Constraint Propagation.
- Automated the… Show more

- Implemented a Decision Tree classifier for predicting credit worthiness of Credit Card applicants based on training data. Over fitting and pruning of the tree was done to handle noise in the input data. Missing data in the query could be predicted from the decision tree built.
- Implemented an automated Sudoku Solver using Min Conflicts algorithm and Backtracking with heuristics which include Minimum remaining values (MRV), Forward Checking and Constraint Propagation.
- Automated the game of Connect Four by implementing Minimax search algorithm with alpha-beta pruning technique in Python. Enhanced features such as Connect-K and Longest-streak-win was also added.
- Implemented iterative deepening search and A-star search (with heuristics) in Python to automate the game of Peg Solitaire.

Technologies: Python
Show less

This project involved implementation of a recommender system by scaling up item-based and user-based collaborative filtering algorithms on Hadoop. An experimental analysis of speedup vs. number of nodes was also done.

Technologies: Python, Hadoop.