Nikhil Podila
Masters graduate from McGill University. I specialize in machine learning applied to complex systems including a Masters thesis applying machine learning on Robotics. I have over 2 years professional experience as a Data Scientist, working with Microsoft Azure Cloud platform, ROS, Spark and PyTorch deep learning frameworks of Tensorflow and PyTorch.
Education:
Master of Science - Thesis
Electrical and Computer Engineering,
Intelligent Systems - Robotics,
McGill University, Montreal, Canada
Advisor: Prof. Hsiu-Chin Lin
Bachelor of Engineering ,
Electrical and Electronics Engineering,
PES Institute of Technology, Bangalore India
Thesis Advisor: Prof. Koshy George
Professional Experience:
Data Scientist/ Associate R&D Engineer,
Robotics & Motion - Drives,
ABB GISPL, Bangalore, India
My Resume 
Robotics
Pick & place + Obstacle avoidance on manipulator & stereocam
  • Utilized Octomap and Moveit for Obstacle avoidance on UR5 Robot with a gripper, while executing Pick & place task.
Avoiding Potential Deadlocks 
  • A method to avoid deadlock of a manipulator when using potential fields for go-to-goal and obstacle avoidance
Trajectory generation - Letters 
  • Trajectory planning using Lyapunov based Dynamical Systems to trace letters from English alphabet
Machine Learning
Linear Classification 
  • Logistic Regression and Linear Discriminant Analysis on UC Irvine datasets - (1) Breast Cancer and (2) Wine Quality
Text Classification 
  • NLP feature engineering and Ensemble Naive Bayes classifiers on Reddit comment classification dataset
Image Classification 
  • Analysis of Convolutional Neural Networks applied on Modified MNIST dataset
Generative Adversarial Networks (GANs) 
  • Survey of GANs on MNIST, FMNIST and CIFAR-10 datasets
Reinforcement Learning
Undergraduate Thesis 
  • Hybrid control of Inverted pendulum (Cart pole) - Swing up (Energy method) and Stabilization.
  • Comparing LQR Stabilization with RL: Tabular and Linear function approximation with Q-learning, SARSA and Actor-Critic
Review of Importance Resampling for Off-Policy Prediction 
  • Reproduced the NeurIPS 2019 paper on Importance Resampling (IR) for Off-Policy Prediction
  • Compared results with Importance Sampling (IS), WIS and Prioritized Experience Replay