Samarth Marudheri

Graduate Student - Robotics

Johns Hopkins University

About Me

I am a graduate student specializing in Artificial Intelligence and Machine Learning at Johns Hopkins University. Prior to graduate school, I worked with Fortune 500 companies to deploy Data Science solutions at scale.

Currently, I am building an open source Python application to learn and teach AI at the JHU Data Science Lab. I am also interning part-time with a Michigan based analytics firm working on Computer Vision problems.

I am passionate about solving real world problems through Data Science tools I have picked up professionally and academically over the years. I also enjoy teaching and have developed case studies in Python/R to implement Statistical, Machine Learning and Deep Learning models. This website showcases my projects and interests.

Interests

Artificial Intelligence
Machine Learning
Data Science

Education

MS in Artificial Intelligence and Robotics, 2019

Johns Hopkins University
Big Data Analytics and Optimization, 2018

Insofe
BSc in Aerospace Engineering, 2015

Embry-Riddle Aeronautical University

Courses

A list of my in-class and online certifications

PGP (Post Graduate Program) in BIG DATA Analytics and Optimization at the International School of Engineering. [The program is certified for quality of content, assessment and pedagogy by the Language Technologies Institue (LTI) of Carnegie Mellon University]
Deep Learning Specialization by deeplearning.ai at Coursera
Mathematics for Machine Learning Specialization by Imperial College London at Coursera
Artificial Intelligence, Machine Learning and Deep Learning by the Computer Science Department at Johns Hopkins University
SQL For Data Analysis at Udacity (Currently pursuing)

Projects

Data Science Case Studies

DeepFake Detection with GANs

The goal of the study is to detect fake images. An autoencoder is trained adversarially to obtain a latent rich representation of real images. The latent representation is then used to reconstruct an image based on the given input image. We propose a comparative study to test if reconstructed features differ strongly for real and fake images leading to enhanced classification performance. We further argue that adversarially training a classifier (DCGAN) generalizes better on unseen data as compared to CNN based architectures.

Swirlypy - A tool to learn and teach AI

The goal of this project is to develop an open-source Python package to learn and teach AI, similar to Swirl for R. It aims to build on the work by Alexander Bauer. Course content and the application are currently under development. Stay Tuned!

Detecting fraudulent insurance claims

The goal of this project is to flag fradulent claims using Computer Vision techniques for an automotive insurance company. The idea is to use misalignment in tables and changes in fonts as signals to classify insurance claims.

Automated Trading Strategy with Reinforcement Learning

In the quest for a robust trading strategy capable of navigating the dynamics of a complex environment, Reinforcement Learning algorithms offer significant advantages over traditional Machine Learning techniques. In this paper, we propose multiple deep learning models capable of predicting signals that capture sentiment and major events that affect the stock prices of companies.

Flight Delay Prediction

Predicted flight delays from structured data with attributes like origin and destination city, departure and arrival times, carrier information, passenger count and weather station data of the cities of interest. Iterated through the Data Science pipeline to build Machine Learning models including Logistic Regression and Decision Trees. Post model building, hosted the app on R Shiny. Welcome to the app!

Resume

Click here to access my resume. (Last updated: June 1, 2020)