Pranoy Kovuri

I am a graduate student at Texas A&M University, where I work on Natural Language Processing and Reinforcement Learning.

I completed my B.Tech (2011-2015) in Electronics and Communication at National Institute of Technology, Warangal

I will be joining from Fall 2019

Email  /  CV  /  LinkedIn  /  Github  /  Facebook

Interests : Machine Learning ■ Deep Learning ■ NLP ■ Reinforcement Learning ■ Distributed Software Development ■ Quantum Computation

Why AI?
The ability to embed intelligence into systems and artificially imitate human behavior, some of which even we are incapable of, is what excites me about AI.
profile photo
Philips Research Cambridge, AI Intern                          2018 Summer
Qualcomm Hyderabad, Software Developer                   2015-2017
Qualcomm Hyderabad, Software Intern                      2016 Summer

For my thesis , working on developing novel NLU technique using semi-supervised technique. Alongside that I am working on solving a critical self-driving problem, when the car interacts with a pedestrian using reinforcement learning.

NLU and Deep Learning, Texas A&M University
  • Developed unsupervised and semi-supervised neural architectures for joint Relation Extraction and NER
  • Modelled and tested sequential architectures for biomedical and sensor time series data
  • Improved performance of XGBoost models on ICU readmission prediction by incorporating text-based features from clinical notes
Reinforcement Learning

Working on a novel research self-driving car problem for pedestrian interaction

  • Developing and testing Q-Learning, DDQN, DDPG and IRL based algorithms for a novel research problem
  • Designing a tailored gym environment for a zebra crossing scenario

Distributed Consensus Raft Algorithm
Team: Niti Jain, Kexin Cui, Ruihong Wang

This work implements a distributed Consensus algorithm called Raft in Python. It keeps the implementation simple to use and easy to adopt.

Github / Report
Sentiment Analysis of Documents

Developed and experimented different neural network architectures for classifying sentiments as positive and negative sentiment based on the movie review. We used IMDb movie review database.

  • Developed a predictive feed-forward neural network for Sentiment Classification for movie reviews.
  • Developed a model based on Recurrent Neural Network (LSTMs) for the same dataset, this takes into account the hidden sequence-related information.
  • Developed a model based on MaxEnt (Logistic Regression). This task was similar to the feedforward task except now we only used one neuron with sigmoid activation functio

Social Networking Web-service for CSE Department
Team: Satya Kesav, Krit Gupta, Sandeep Gottimukkala, Kevin Mathew

This is a Web application dashboard for TAMU CSE Students, Student Bodies and Administrators, that we made as part of the course CSCE 606. The team members were: Satya Kesav, Krit Gupta, Pranoy Kovuri, Sandeep Gottimukkala and Kevin Mathew (Team CosmicSoft).

The project was developed in a complete Agile environment, with multiple iterations, each encompassing a short product demo (with completed features) to the client, TDD/BDD, regular and proper documentation. We achieved a test coverage of over 93%!

Technologies/Frameworks/Languages used : Ruby on Rails, Heroku, Cloud9, PostgreSQL, Bootstrap, Pivotal Tracker, Git, JavaScript, SimpleCov, Capybara, Cucumber.

Website / Youtube / Documentation / Github

Website Template Credits: Jon Barron