Project-Flume: Realtime Flood warning system

I built a real-time environmental monitoring platform processing 14GB daily from 2,600+ Canadian weather stations, implemented ML-based flood prediction, and deployed it on AWS with Infrastructure as Code.

Portfolio Allocation with Financial Reinforcement Learning

This project leverages PPO reinforcement learning to optimize stock portfolios, achieving a 21.97% return with a 0.63 Sharpe ratio and controlled 12.02% maximum drawdown. The complete system includes adaptive allocation, hyper parameter tuning, and stress testing to ensure resilience accros markets.

Air Quality Index Forecasting

Developed predictive models for AQI forecasting using Time series algorithms like ARIMA, XGBoost, and LSTM, achieving 90% accuracy with ARIMA for the next 60-day forecast. Analyzed and preprocessed large datasets with over 1 GB of size to identify trends and seasonal patterns in pollutants.

Product Recommendation using Sentiment Analysis

Developed an End-to-End Product recommendation system. Implemented NLTK for clustering unlabelled data and used python, SQL for data processing. Employed machine learning algorithms with pipelines for accuracy and sentiment score is calculated using Vader. Visualized trends using Matplotlib.

App Rating Prediction

This Python program uses the Pandas and NumPy libraries for data manipulation, Seaborn, Matplotlib for data visualization, and scikit-learn for linear regression modeling. The program follows several steps for data preprocessing, cleaning, analysis, and regression modelling on a Playstore dataset.

House Sales Dashboard

Provides stakeholders, real estate agents, and prospective buyers with a user-friendly platform to analyze, understand, and derive insights from house sales data. Supports data-driven decision-making by offering a visually appealing and interactive interface.

Car Price Prediction

The primary goal is to predict the selling price of cars based on various attributes, helping understand the factors that most influence car prices. XGBoost Regressor appears to be the most effective, followed by Decision tree regressor and Linear Regression.