Vasanth Reddy

Experience

• 

  Department of Science and Technology

  Oct 2017 - May 2018

  Student Research Assistant

  Used Sequential Machine Learning models such as Hidden Markov Models and Deep Learning models, Recurrent Neural Networks and Long Short Term Models(LSTM) to forecast wind speed. In processing the data we used Wavelet decomposition method to increase the accuracy of the model. We obtained significant results using these algorithms.

• 

  Infomatics Corp

  May 2020 - Aug 2020

  Data Science Intern

  - Built ML models using various algorithms in the Dataiku DSS platform: Responsible for creating predictive models in DataIKU Data Science Studio (DSS) using a variety of algorithms- Created architecture flow for model implementation: Designed the workflow and structure for implementing these models within the system. Implemented preprocessing techniques (visual, code, visual ML recipes): Applied data preprocessing methods, including visual and code-based techniques, to prepare the data for modelling.- Implemented predictive analytics, clustering, and prediction models: Applied supervised learning methods, clustering techniques, and predictive models to address specific business needs within Dataiku DSS Show less

• 

  Virginia Tech

  Aug 2020 - May 2021

  - Proposed a novel mechanism based on the heat diffusion mechanism to enable explicit parameter sharing across agents in a dynamic environment. - We introduced multiple channels with heat kernels in parallel to capture different variations of information, enhancing the information-enriched encoder. CS 5824/ ECE 5424 - Advanced Machine LearningCS 4984 - Machine Learning Capstone Project

  • Graduate Research Assistant

    Dec 2020 - May 2021

  • Graduate Teaching Assistant

    Aug 2020 - May 2021
• 

  Siemens

  May 2021 - Sept 2021

  Research Intern

  - Developed a modular framework for Multi-Agent extension of off-policy RL algorithms (DQN, DDPG, TD3, and SAC) and on-policy RL algorithm (PPO) and their variants, resulting in improved performance and code reusability.- Implemented direct MARL algorithms, including Multi-Actor-Attention-Critic (MAAC), VDN, QMIX, and FACMAC, for tackling complex coordination problems in diverse environments.- Evaluated the developed algorithms against benchmark algorithms on environments such as MPE, Multi-Agent MuJoCo, Traffic signal control, and Power grid systems, demonstrating superior performance and scalability. Show less

• 

  Virginia Tech

  Sept 2021 - now

  - Worked on a project that leverages satellite imagery for detecting heavy metal-induced stress in agroecosystems, crucial for ensuring food security. - The project involved processing Sentinel-2 satellite images spanning four years and calculating various Vegetation Indices. We developed an interactive Python program to analyze index time-series data, addressing challenges like non-stationarity and missing data through interpolation.- We applied clustering methodologies, including Hierarchical DL Methods and K-Means Clustering, to group locations with similar index patterns and uncover potential irregularties. Show less

  • Graduate Teaching Assistant

    Aug 2023 - now

  • Graduate Research Assistant

    Sept 2021 - Jul 2023