# Projects And Technical Papers


# Projects

## Heart Disease Prediction Model
- Objective: Develop and deploy a machine learning model to predict heart disease based on indicators
- Technologies:
  - Frontend: Streamlit
  - Backend: FastAPI and Docker
  - Experiment Logs: MLFlow on DagsHub 
  - Cloud Services: Digital Ocean
- Key Features:
  - Normalized database creation and data exploration
  - Preprocessing pipelines and multiple classification algorithms with feature engineering, selection, and dimensionality reduction (PCA)
  - Model deployment with FastAPI, Docker, and Streamlit
  - Comprehensive documentation and project presentation via Jupyter Book
- Link: 
  - [Streamlit Frontend](https://apprender-ghfqkbcthxbxrfdgcomw23.streamlit.app/)
  - [Description](https://varunnn789.github.io/Heart-Disease-Prediction/index.html)

## Restaurant Database
- Objective: Full-stack restaurant website with user authentication and reservation system with dynamic menu
- Technologies:
  - Frontend: HTML5, CSS3, JavaScript
  - Backend: Node.js, Express.js
  - Database: PostgreSQL
  - Authentication: JWT, bcrypt
  - Cloud Services: Render, Koyeb
- Key Features:
  - User authentication system with secure login/signup
  - Dynamic menu management with database integration
  - Real-time reservation system with user tracking
  - RESTful API endpoints for data management
- Link: 
  - [Restaurant Website](https://restaurant-frontend-gn13.onrender.com/#reservations)
  - [Description](https://varunnn789.github.io/restaurant-website/)

## Crime Detection using Sentiment Analysis
- Objective: Develop an innovative crime prediction model integrating sentiment analysis from social media with historical crime data
- Technologies:
  - Data Collection: Playwright, PRAW
  - Data Processing: Pandas, NLTK, TextBlob
  - Machine Learning: Scikit-learn (PCA, Random Forest)
  - Data Visualization: Matplotlib
- Key Features:
  - Automated extraction of crime-related social media data from Twitter and Reddit
  - Custom sentiment scoring mechanism with crime severity weighting
  - Integration of real-time sentiment analysis with historical crime data
  - Principal Component Analysis (PCA) for feature selection and dimensionality reduction
  - Random Forest classification for crime trend prediction
  - Geolocation-based crime severity mapping at the state level
  - Comparative analysis of predicted crime severity rankings against established crime indexes
- Link:
  - [Description](https://varunnn789.github.io/Crime-Detection-Using-Sentiment/)

## Generative AI for Mental Health Support
- Objective: A generative AI-powered web application to provide therapeutic suggestions, mindfulness exercises, and empathetic responses for mental health support.
- Technologies:
  - Frontend: React.js, HTML5, CSS3
  - Backend: Flask, FastAPI
  - AI Models: OpenAI GPT-4, Hugging Face Transformers
  - Database: MongoDB
  - Cloud Services: AWS, Render
  - APIs: Twilio (optional for voice/text), OpenAI API
- Key Features:
  - Emotion-based suggestions for therapeutic exercises and motivational messages
  - User-customizable tone and response preferences
  - Privacy-focused design with no data storage
  - Safeguards to redirect users to professional resources when necessary
  - Multimodal support, including text and optional audio-guided exercises

## NarrativeQA Reading Comprehension
- Objective: Produce natural language results from summaries and questions
- Technologies: BiDAF attention model, GRU, LSTM, RNN
- Key Features:
  - Natural language processing
  - Advanced attention mechanisms

## Machine Learning for Indoor Navigation
- Objective: Assist vision-impaired individuals with indoor navigation
- Technologies: Reinforcement Learning, DeepMind lab, Q-learning algorithm
- Key Features:
  - Simulated agent navigation
  - Reward and penalty calibration


# Technical Papers

## Navigating Uncertainty: Evaluating Risks to Enhance Drug Sales Forecast
- The paper highlights the use of Monte Carlo Simulation to tackle uncertainties in new drug sales forecasting within dynamic market conditions.
- It explores integrating Monte Carlo with AI/ML for enhanced probabilistic forecasting and future advancements.
- Link: 
  - [Journal of PMSA Spring 2024](https://www.pmsa.org/_resources/documents/journal/Journal-of-PMSA-Spring-2024.pdf)