My Software Solutions, Live Projects, and Libraries
Welcome to my portfolio of projects. For more, visit my GitHub profile.
M.S. Thesis: Blindfolded Spider-Man Optimization
Topic: Blindfolded Spider-Man Optimization: A Single-point Metaheuristic Suitable for Continuous and Discrete Space
- Developed a new metaheuristic algorithm based on Buggy-Pinball using high dimensional geometric math up to 20D.
- Tackled higher dimensional non-convex optimization problems.
- Applied metaheuristics to unbounded knapsack discrete problems with implementations ranging from 2D to 20D.
- Designed a neural network in PyTorch to implement heuristic algorithms as optimizers while initializing neural network weights.
- Conducted a comparative analysis of benchmark meta-heuristics for complex optimization problems.
- Analyzed data for statistical significance tests like Shapiro-Wilk, Kruskal-Wallis, and converse test with a 0.05 p-value, alongside Q-Q plots for dataset normalization.
-
Researched LLM’s biases on sentiment analysis of news feeds on world crises.
- Read more about Buggy Pinball
- My Thesis work - Dissertaion
- GitHub Repository - The repo is private as the work is not published yet.
Tugposts.com
Visit: Tugposts
- Developed as a one-stop online resource with extensive data compilation.
- Implemented as a Spring Boot REST application following MVC architecture.
- Designed the static HTML using JavaScript and opted for MySQL over NoSQL based on project requirements.
- Deployed on AWS using EC2, RDS, VPC, IAM, and NginX for reverse proxy setup.
Machine Learning Projects
Large Language Models (LLM)
Exploring the capabilities of large language models, this project focuses on text completion and summarization. The model provides personalized insights by processing user-uploaded CSV data.
- Features: Text completion, summarization, and personalized insights.
- Technology: Built on the Langchain framework.
- User Interaction: Users upload data via a local system path to receive insights.
- Technologies: LLM, Langchain, Llamaindex, VectorDB, Huggingface libraries, ML Pipelines (Ingestion-Train-Inference), PyTorch, Transformer Architecture, Vector-Embeddings, Streamlit etc ect…
- Note: This is an ongoing open project work and it will be updated regularly with new technologies. -GitHub Repository
Data Analysis and Text Summarization (DATS)
This project implements text completion and summarization capabilities of LLMs to provide insights from personal data. Users can upload CSV files and ask for specific insights, such as job application statuses for a particular month.
- Implementation: Utilizes text analysis and summarization to provide insights from CSV data.
- Purpose: A personal tool for understanding and utilizing the Langchain framework.
- Example Query: “Provide me insights on my job applications in March 2024.”
Fake News Detection
Goal was to predict whether news is real or fake, utilizing Python’s sci-kit and nltk libraries with various NLP techniques and machine learning algorithms.
GitHub Repository - This is one of many small implementations of machine learning projects.
University ListServe
A ListServe project developed using multiple design patterns like composite, proxy, observer, factory, and singleton for a university’s notification, message delivering, and SOS system.
Custom WC Command Line Tool
Built a custom version of the Unix command line tool wc to calculate lines, word count, byte size, and character count for input files.
- Supports reading from standard input if no filename is specified.
- Extended functionality through factory design, allowing further expansion.
Custom Compression Tool like Unix
Developed a tool for compressing and decompressing text files using Huffman coding, including bit manipulation for header creation and reading.
- Commands include “-c” for compression and “-d” for decompression.
- Utilized factory and singleton design patterns.
Data Structure and Algorithm Practice
An open repository created to practice data structure questions from platforms like Leetcode and Geeks for Geeks.
- Includes solutions from external resources and open-source communities.
Each project showcases not just my technical skills but also my passion for creating and contributing to innovative solutions in the field of software development.