Projects & Portfolio
Showcasing my technical expertise through innovative solutions
Aws Iam User Management
Completed
AWS IAM User Management Tool is a Python-based command-line utility for managing AWS IAM users efficiently. It allows administrators to automate common IAM tasks such as creating users, attaching policies, generating access keys, and cleaning up unused credentials.
Key Features
- Automated IAM Setup: Creates users (Alice, Bob, Charlie, Dana), groups (Developer, DBAdmin, Auditor), and policies for S3, EC2, RDS, and CloudTrail access.
- Policy Validation: Checks for overly permissive policies (e.g., Action: *, Resource: *).
- Cleanup Script: Safely deletes all created IAM resources.
- Unit Tests: Uses pytest and moto to test setup logic without AWS API calls.
- Configurable: Uses config.yaml for customizable team and permission settings.
Topics
access-management
aws
aws-cli
aws-iam
aws-tools
boto3
devops
iam
identity-and-access-management
user-management
Completed
Jul 2, 2025
Linkedin Auto Apply
Completed
Automate your LinkedIn job applications with ease! This Python-based bot uses Selenium to search for jobs and automatically apply to positions with Easy Apply. Designed to save time and streamline your job hunt while mimicking human-like behavior to reduce detection risk.
Key Features
- Automates LinkedIn job applications using Selenium.
- Searches for jobs based on user-defined keywords and filters.
- Automatically applies to positions with 'Easy Apply'.
- Mimics human-like behavior to reduce detection risk.
- Configurable application settings (resume, cover letter, job preferences).
- Saves time and streamlines the job application process.
- Logs applied jobs and application status for tracking.
- Handles multiple job applications in a single session.
Topics
automation
linkedin
linkedin-bot
selenium
Completed
Jun 10, 2024
Chinese License Plate Recognition
Completed
Implementation of Chinese license plate character segmentation and recognition system
Key Features
- Implementation of both traditional and deep learning methods for license plate recognition.
- Preprocessing of dataset images using OpenCV for color conversion, noise removal, and perspective transformation.
- License plate detection using morphological operations, contours, and aspect ratio filtering.
- Perspective transformation to obtain front-view license plate images.
- Character segmentation and template matching for traditional recognition.
- Background checking for different plate colors (blue, yellow, green, white).
- CNN-based deep learning model trained on individual character images for high accuracy.
- Use of CCPD dataset with bounding box and character-level annotations for deep learning.
- Training and evaluation with batch normalization, max pooling, and multiple hidden layers.
- High accuracy achieved: 99.75% on 2000 images and 99.94% on 60,000 images.
- Configurable preprocessing pipeline to handle slanted or skewed plates.
- Supports both single-character and full license plate recognition.
- Potential for augmentation to improve performance on hazy or obstructed images.
- Comparison between traditional and CNN-based approaches for evaluation.
Topics
computer-vision
convolutional-neural-networks
image-processing
license-plate-detection
license-plate-recognition
opencv
Completed
Nov 29, 2022
WikiSearch
Completed
WikiSearch is a search engine built on Wikipedia dump datasets, designed to help users quickly find relevant articles. It supports fuzzy search, wildcard queries, and multiple ranking methods to improve search accuracy and user experience.
Key Features
- Search engine built from Wikipedia dump dataset (15M+ articles).
- Supports document vector, inverted, positional, and bi-gram indexing.
- Fuzzy search with spelling correction using Levenshtein and bi-grams.
- Wildcard search for partial or unknown terms.
- Five ranking algorithms: Bag of Words, Cosine, Jaccard, TF-IDF, High-IDF.
- Preprocess XML dumps into JSON chunks for fast access.
- Optimized query processing with precomputed indexes and scoring.
- User-friendly GUI with search bar, ranking selection, and suggestions.
- Displays top-k results with title, link, and snippet.
- Handles large datasets efficiently with high accuracy.
- Query performance measured for single and multi-term searches.
- Extensible design for new algorithms, preprocessing, and UI features.
- Corrects typos and handles partial or slanted queries.
- Future improvements: better fuzzy weighting and precomputed TF-IDF.
Topics
fuzzy-search
indexing
python
reactjs
search
search-algorithms
search-engine
searching-algorithms
wikipedia
wikipedia-dump
Completed
Apr 30, 2021
One Shot Learning
Completed
One-shot Image Classification leverages deep learning to classify images with very few examples per class, using Siamese Neural Networks and One-shot Learning techniques.
Key Features
- Supports One-shot Learning for image classification
- Uses Siamese Neural Network for similarity comparison
- Handles small labeled datasets efficiently
- N-way testing for model validation
- Triplet Loss implementation for improved embeddings
- Supports Omniglot dataset for character recognition
- 1-Nearest Neighbor and Random baseline models included
- Built with Keras and Python for deep learning experiments
- Generates similarity scores between image pairs
- Can be applied to face recognition, signature verification, and other small-data tasks
Topics
computer-vision
deep-learning
few-shot-learning
image-classification
machine-learning
omniglot-dataset
one-shot-learning
python
siamese-neural-network
signature-verification
Completed
Apr 27, 2021
Image Stitcher
Completed
Python project for computing image homography, drawing correspondences, warping and blending images, and generating realistic billboard overlays using OpenCV and NumPy.
Key Features
- Interactive point selection for computing homography
- Compute homography matrix from source and destination points
- Draw correspondences between images
- Warp and blend images seamlessly using computed homography
- Replace billboard content realistically with a custom image
- Supports multiple image formats (JPG, PNG)
- Uses OpenCV and NumPy for efficient image processing
- Saves results including warped images, correspondence maps, and billboards
- Handles edge cases and checks for minimum required points
- Fully modular functions for integration into larger projects
Topics
augmented-reality
computer-vision
image-blending
image-processing
image-stitching
image-warping
opencv
python
Completed
Apr 22, 2021
Harris Corner Detector
Completed
Implementing Harris Corner Detector and Scale-Invariant Feature Transform (SIFT) for keypoint detection, feature description, and image matching under transformations like rotation, scaling, and translation.
Key Features
- Custom implementation of Harris Corner Detector
- Corner response calculation with Gaussian smoothing and Sobel filters
- Keypoint orientation and scale estimation
- Patch-based feature descriptors with normalization
- Feature matching using Euclidean distance and DMatch
- Evaluation of rotation, translation, and scaling transformations
- Visualization of matched features between two images
- SIFT-based feature detection and matching with BFMatcher
- Comparison between Harris + custom descriptors vs. SIFT
- Automatic saving of output results for analysis
Topics
computer-vision
feature-detection
harris-corners
image-matching
image-processing
keypoint-matching
opencv
pattern-recognition
python
sift
Completed
Apr 1, 2021
DIP Super Resolution
Completed
An efficient deep learning approach for Single Image Super-Resolution (SISR) that leverages dense projection networks with up- and down-sampling stages, iterative error feedback, and deep concatenation.
Key Features
- Designed and implemented Dense Projection Network (DPCNN) for efficient LR → HR image mapping
- Applied Iterative Error Feedback to progressively correct reconstruction errors
- Developed multi-stage up & down sampling for robust feature learning
- Utilized deep concatenation to enhance reconstruction quality
- Built multi-channel (RGB) support for high-fidelity restoration
- Achieved state-of-the-art performance, outperforming SRCNN, FSRCNN, VDSR, DRRN, and EDSR
- Optimized system for lightweight, fast, real-time applications
- Conducted benchmark evaluations on Set5, Set14, RAISE 1K, BSDS300 datasets
- Implemented in PyTorch, enabling an extensible training and inference pipeline
- Explored applications in security, medical imaging, and photography enhancement
Topics
computer-vision
deep-learning
dense-projection-networks
image-enhancement
image-processing
image-restoration
image-super-resolution
python
pytorch
single-image-super-resolution
sisr
super-resolution
Completed
Dec 19, 2020
LC4 Binary Interpreter
Completed
C-based LC4 interpreter and disassembler that executes LC4 binaries, manages memory, and converts object files into readable assembly for educational use in computer architecture.
Key Features
- Binary Interpreter: Executes LC4 machine code instructions from compiled object files
- Disassembler: Converts LC4 binary into human-readable assembly instructions
- Memory Management: Implements a simulated LC4 memory system for instruction and data handling
- Loader Module: Loads object code (.obj) files into LC4 memory for execution
- Instruction Execution: Supports decoding and execution of LC4 instruction set architecture
- PennSim Integration: Compatible with PennSim LC4 environment for running and testing programs
- Testing Suite: Includes test files and scripts for automated validation (makefiletest.sh)
- Documentation: Includes project.pdf explaining implementation details and design decisions
Topics
assembly
binary-interpreter
c-language
emulator-compiler-design
lc4
lc4-interpreter
lc4-simulator
pennsim
Completed
Aug 19, 2020
Bangla Newspaper Scraper
Completed
A Python scraper to extract articles from popular Bangla newspapers, supporting text extraction, link collection, and data storage.
Key Features
- Scrapes articles from multiple Bangla newspaper websites
- Extracts headlines, article text, publication date, and links
- Handles pagination and 'load more' buttons
- Saves scraped data in structured formats (CSV/JSON)
- Supports scheduling for regular scraping
- Lightweight and easy to extend for additional newspapers
- Includes basic data cleaning and preprocessing
- Optimized for performance and reliability
Topics
automation
bangla-news-scraper
beautifulsoup
data-collection
python
selenium
Completed
Feb 11, 2019
Enemy Shooting Game
Completed
A 2D enemy shooting game built using OpenGL and C++. Includes interactive shooter, moving enemies, scoring, lives, and a high score system.
Key Features
- 2D top-down shooting game built with C++ and OpenGL
- Interactive shooter controlled using arrow keys
- Shoot bullets using spacebar to destroy enemies
- Enemies spawn dynamically with varying speed and position
- Score tracking and high score system with file storage
- Life system – lose life if enemies pass the shooter
- Main menu with New Game, High Score, How to Play, and About sections
- Real-time collision detection between bullets and enemies
- Visual effects using OpenGL primitives (spheres, polygons, lines)
- Game over screen with score display and restart option
Topics
computer
computer-graphics
computer-graphics-opengl
cpp
game-development
graphics
opengl
shooting-game
Completed
Dec 22, 2017
Smart Audio Profile Manager
Completed
A mobile app that automatically adjusts device audio profiles based on position and user activity using sensors like accelerometer and proximity.
Key Features
- Automatically switches audio profiles based on device position
- Uses accelerometer and proximity sensors to detect device orientation
- Silent mode when the device is upside down
- Vibration mode when the device is in a pocket
- General audio profile when the device is in hand or on desk
- Customizable audio profiles according to user preferences
- Ability to create custom profiles for specific situations like driving or meetings
- Seamless and automatic audio management without manual intervention
- Lightweight and battery-efficient implementation
- User-friendly interface for profile management
Topics
accelerometer
android-application
android-audio-profile
android-sound-automation
context-aware-audio
gyro-sensor-audio
motion-based-audio
proximity-sensor
Completed
Apr 7, 2017
Hangman Console Game
Completed
A console-based Hangman game in C# supporting single-player and multiplayer modes.
Key Features
- Console-based Hangman game implemented in C#
- Supports single-player mode against the computer
- Supports multiplayer mode for two players
- Tracks guesses and displays correct/incorrect letters
- Visual representation of Hangman progress
- Customizable word lists
- Simple and interactive command-line interface
- Handles input validation and repeated guesses
- Score tracking for multiplayer games
- Lightweight and easy to run without additional dependencies
Topics
cli
console-game
csharp
hangman
Completed
Oct 14, 2016
Tower Of Hanoi Game Java
Completed
A Java Swing implementation of the classic Tower of Hanoi puzzle game with interactive graphical interfaces.
Key Features
- Implemented in Java using Swing for graphical interface
- Classic Tower of Hanoi puzzle game
- Objective: Move all disks from Pole A to Pole C following the rules
- Rules enforced: Cannot move larger disk onto a smaller disk
- Calculates minimum moves required: 2^n - 1
- Supports any number of disks
- Interactive GUI gameplay with drag-and-drop functionality
- Console interface also supported for traditional input
- Visual and audio feedback for moves
- Educational tool for understanding recursion and problem solving
Topics
game
java
java-swing
puzzle-game
recursion
swing
tower-of-hanoi
Completed
May 16, 2016