Global competition by Amazon Web Services to build autonomous driving vehicle

Self Driving Car Development | Won the global Financial Services Event for fastest self-driving car
•Used AWS SageMaker for training model used by car for completing the race lap autonomously
•Used EDA techniques for log analysis of training data to improve model & tune hyperparameters
•Overcame the challenge of reward function sniffing out an unforeseen behavior and improving it

Twitter tweet Sentiment Analysis | Won the competition among 20 teams for best ML pipeline
• Used advanced NLP techniques for data preprocessing & various algorithms for model training
• Built end to end pipeline for preprocessing, model training and performance measurement
• Extracted entities towards which sentiments were directed at using KNN, RNNs, LSTMs for NER

Competition for various Regression and Classification tasks using Data Science techniques

Foundational Program of 5 courses: NN in TensorFlow | Understood the capabilities, challenges, & consequences of deep learning

• Neural Networks & Deep Learning: Built and trained deep neural networks, implemented vectorized neural networks, identified architecture parameters, and applied Deep Learning to various applications

• Improving Deep Neural Networks: Hyperparameter Tuning, Regularization & Optimization: Used best practices to train & develop test sets, analyzed bias/variance for building DL applications and applied optimization algorithms

• Structuring Machine Learning Projects: Used strategies for reducing errors in Machine Learning systems, understood complex Machine Learning settings, and applied end-to-end, transfer and multi-task learning

• Convolutional Neural Networks: Built a Convolutional Neural Network, applied it to visual detection and recognition tasks, used neural style transfer to generate art, and applied these algorithms to image, video, and other 2D/3D data

• Sequence Models: Built and trained Recurrent Neural Networks, its variants (GRUs, LSTMs), applied RNNs to characterlevel language modeling, used Word Embeddings, tokenizers and transformers to perform Named Entity Recognition