Auto ML

Welcome to my user-friendly Machine Learning (ML) model creation platform designed for individuals with limited or no prior ML experience. This platform empowers users to effortlessly create ML models through an intuitive interface, drag-and-drop functionality, and pre-built templates for various ML tasks such as classification, regression, and clustering.

Automated Cuet Score Checker

In CUET, checking scores manually can be a tedious task, involving matching each answer with the original sheet. Introducing the Automated CUET MCA Score Checker - a convenient solution to simplify the score-checking process.

OpenAI Whisper Automated Speech Recognition(hindi)

This is a project focused on building a robust speech recognition system for the Hindi language. It leverages OpenAI's Whisper model to convert spoken Hindi into accurate text, facilitating applications in transcription, voice commands, and more. This project aims to improve accessibility and efficiency for Hindi speakers in technology-driven contexts.

Sentiment Analysis using BERT

• Utilized BERT, a leading NLP model, to conduct accurate sentiment analysis.
• Contextual Understanding: Leveraged BERT's contextual awareness to capture nuanced meanings, surpassing conventional methods.
• Actionable Insights: Provided valuable insights into customer feedback and social sentiment, facilitating data-driven decision-making processes.

Transformer-based (Decoder-only) Language Model from Scratch

This repository explores building a character-level transformer decoder in PyTorch, similar to GPT while focusing more on understanding individual components. My goal is to gain deep transformer knowledge and see if character-level learning improves handling of unseen words. The code allows for hyperparameter tuning and experiment customization.

Document QnA using Llama3 and Groq

Document QnA is a webapp that lets users upload multiple documents and ask questions about their content. It uses Llama3, Groq API, LangChain, FAISS, and Google Palm Embeddings to identify relevant documents and provide answers with page numbers. The Streamlit interface ensures easy and efficient use.

Blog Generation using Llama2

• Developed a custom blog generator using Llama2, Langchain, and Huggingface.
• Designed a Streamlit interface allowing users to specify topic name, world count, and audience type for personalized blog content.

Movie Name Guessing using Plot with RAG

An AI-powered application that can guess movie titles based on plot summaries. Built using LangChain, Google Palm LLM, CSVLoader, RetrievalQA, Google Palm Embeddings, and FAISS. Deployed on Streamlit for an interactive user experience, allowing you to enter a plot summary and receive a predicted movie title.

Chat with databases using RAG

• Deployed an end-to-end RAG application with a monitoring system by utilizing LangChain and LangSmith.
• Leveraged advanced language models like GooglePalm for natural language understanding.
• Employed ChromaDb for database interaction, providing a rich palette of functionalities.
• Optimized query generation using FewShotPromptTemplate and create_sql_query_chain.

House Price Prediction with experimentation Tracking using MLFlow

• Developed a house price prediction model using linear regression, predicting prices based on features like location, total square feet, number of bathrooms, and bedrooms (BHK).
• Experiment Tracking: Implemented MLflow for tracking experiments, streamlining the development process and providing deeper insights into model performance.
• Hyperparameter Tuning: Utilized GridSearchCV and Hyperopt to fine-tune models, optimizing performance and achieving better results.

Titanic Survival Prediction

Using machine learning techniques, I developed a model to predict survival rates aboard the Titanic. By analyzing passenger data, I gained insights into factors influencing survival, contributing to disaster preparedness strategies.

Iris Flower Classification

Implemented a classification model to accurately classify Iris flower species based on their features. This project honed my skills in data preprocessing, model selection, and evaluation, demonstrating proficiency in classification algorithms.

Credit Card Fraud Detection

Developed a robust fraud detection system using machine learning algorithms to identify fraudulent transactions. Leveraging techniques like anomaly detection and feature engineering, I contributed to enhancing financial security measures.

Myfiglet

• Created a Python library enabling effortless integration and display of FIGlet fonts within Python CLI applications.
• This is a simple and self-contained library that doesn't require external dependencies. It enhances the visual appeal of various programs and enables the addition of colors to figlet fonts with the color parameter.

Py4Math

• Created a Python library to streamline and enhance coding workflows.
• Highlighted by a powerful search() function, users can instantly resolve queries, with the module currently featuring over 50 mathematical formulas and ongoing development.