MNIST Digits Recognition

This project focuses on recognizing handwritten digits using a deep learning model trained on the MNIST dataset. The model is deployed as a REST API using FastAPI, packaged into a Docker container and orchestrated using Kubernetes for scalable deployment. The API allows users to upload digit images and receive predictions in real time.

• Check it out

Multi Agent Audio Tour System: AI Audio Tour Guide

Developed a Multi Agent Workflow for an AI Tour Guide using OpenAI Agents SDK and OpenAI TTS. The application generates location-specific content tailored to user interests and converts it to customizable tone audio tour. A user-friendly Streamlit interface allows users to select locations and preferences for an immersive experience.Orchestrator agent manages the workflow among various agents. Websearch integrated in every agent for fetches real-time data, Content Generation which creates engaging narratives and Text-to-Speech Conversion which transforms text into natural-sounding audio.

• Check it out

Pharmaceutical Insight Retrieval System

Developed an advanced Retrieval-Augmented Generation (RAG) pipeline with similarity search for precise query responses. It is designed to help users gain meaningful insights from research papers and documents in the pharmaceutical domain. It uses LangChain, Google Gemini, ChromaDB and Streamlit.

• Check it out

Boa AI

• Coming soon...

Vision AI

Visually impaired individuals often face challenges in understanding their surroundings. Our project aims to address this by providing:

• Real-time scene analysis
• Text-to-speech conversion
• Object and obstacle detection
• Personalized assistance for tasks

• Check it out

Diamond Price Prediction

Developed a machine learning model to predict diamond prices based on features like carat, cut, color and clarity. Automated MLOps pipeline with stages for Data Ingestion, Validation, Transformation, Model Training, Evaluation and Prediction. Deployed the app on AWS EC2 for production use.

• Check it out

Text Summarizer App

Tech Stack: Python, Transformers, NLP, MLOps, Docker, CI/CD, AWS

• Developed a text summarizer app that will summarize any text, dialogue, conversation, or article.
• Utilized the Pegasus model and fine-tuned it with the SAMSum dataset.
• Implemented an automated MLOps pipeline with stages for Data Ingestion, Validation, Transformation, Model Training, and Evaluation to ensure a smooth workflow.
• Built a Prediction Pipeline for Model Serving using a Streamlit app.
• Integrated an automated CI/CD pipeline using GitHub Actions.
• Deployed the application on AWS EC2 via ECR for production use.

• Check it out

Wine Quality App

Tech Stack: Python, MLFlow, Docker, CI/CD, Flask, AWS

• Developed a web app to check wine quality on a scale of 0-10.
• Implemented pipelines for Data Ingestion, Validation, Transformation, Model Training and Model Evaluation
• Used MLFlow for experiment tracking, Docker for containerization, and GitHub Actions to automate CI/CD
• Created a Flask UI, deployed on AWS EC2 via ECR

• Check it out

Automatic Language Detection

Tech Stack: Python, NLP, Classification, Docker, FastAPI

• Developed a language detection system capable of identifying any language.
• Performed NLP, creating a bag of words using CountVectorizer.
• Achieved 97% accuracy using a Naive Bayes classifier.
• Created a web API with FastAPI and deployed on the Render Platform.

• Check it out

Spotify Recommendation System

Skills: Python 3, Pandas, NumPy, Scikit-learn, FastAPI

• Built a Spotify recommendation system tailored to individual preferences. Users can input their favorite song and year to receive music recommendations.
• Performed tasks such as Feature Engineering, Exploratory Data Analysis (EDA), data fetching using Spotify Web APIs, and Model development and serving using FastAPI.

• Check it out

Anime Recommendation System

In this project, I've built an anime recommendation system to individual preferences. To achieve this, I performed various tasks, including Data cleaning, Data preprocessing, Feature engineering, EDA, Content-based filtering and Model Evaluation.

• Check it out

Reinforcement Learning for Mario Game

Super Mario Bros. game, which was released in 1985 for the Nintendo Entertainment System, had 32 levels spanning across 8 worlds. The Mario Agent successfully cleared 6 levels from these 8 worlds.

• Check it out