Shreya Banik

Developed an AI agent capable of autonomously playing the Snake game by implementing reinforcement learning algorithms.
Employed Pygame library for the game environment and Pytorch for implementing a suitable RL algorithm to train the agent for optimal decision making.
Adjusted QTrainer for the neural network weights to improve Q-value predictions and plotted Visualization to track the improvement in training performance by checking on average score by time.

Developed and implemented a predictive model for Uber ride fares, leveraging diverse input features using Machine Learning.
Identified and managed outliers in critical features, making informed decisions on their removal or application of appropriate transformations.
Trained the data model on a 1.6M parameter dataset and efficiently employed a Random Forest Regressor to split data for training/testing, scale numerical features.
Optimized performance using metrics such as Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) resulting in an 80% increase in training efficiency.

Led end to end development of an adult income prediction system, utilizing Microsoft Azure and Machine Learning concepts, achieving 94.4 %.
Performed data cleaning to handle null values, feature se lection, and class imbalance via up sampling, training two models for comparison.
Achieved AUC of 0.986 in boosted decision tree model after hyperparameter tuning, published as a web service for inference.

Analyzed a dataset of 51.3K entries using Python and key libraries (Numpy, Matplotlib, SciPy, Pandas, Seaborn) to uncover crucial insights into the alarming increase in accidental drug-related deaths in Connecticut (2012-2022).
Applied advanced statistical analysis to map and summarize evidence, providing a profound understanding of changing patterns in drug-related deaths, contributing to public health surveillance and breaking down complex dynamics through the incorporation of toxicology results.
Established a foundation for effective interventions, empowering public health authorities with valuable insights to potentia lly achieve a 30% reduction in drug related fatalities in the community through informed policy decisions.

Spearheaded development of a Stroke Prediction Model with R programming and r-libraries (dplyr, tidyr, caret, ggplot, tidyverse etc.) in R-studio.
Deployed the validated linear regression model under Machine Learning concept for clinical use, enhancing decision-making in healthcare services.
Visualized data to analyze the primary factors contributing to the phenomenon behind stroke.
Conducted comprehensive data analysis, preprocessing, and feature selection to achieve 94.62% predictive accuracy.

This objective is to explore the domain of Internet of Things (IoT) is used mainly in Defense and Rescue purpose on sea surface.
This USV will provide better security and surveillance capabilities to our Sea borders and restrict unauthorized infiltration attempts.It could also be utilized to save lives in case of a natural disaster.
It has an onboard GPS for Location tracking and in case of any anomaly Quick response team can be sent at that location.