Game-Playing Agent | view on GitHub
Designed a game-playing agent using Adversarial search techniques, Minimax and AlphaBeta, as well as custom evaluation functions to play Isolation effectively and efficiently. Isolation is “a deterministic, two-player game in which the players alternate turns moving between cells on a regular grid till one of them has no more moves left.”
Stacks Image 625
Planning Search | view on GitHub
Solved Air Cargo planning problems defined in PDDL using various search techniques, including standard search algorithms such as Breadth First Search, as well as A* search. A* search used domain-independent heuristics obtained by analyzing a Planning Graph -- also implemented as part of the Planning Search project.
Stacks Image 633
Sign Language Recognition System | view on GitHub
Recognized gestures in American Sign Language using Hidden Markov Models, relied on model selectors such as Bayesian Information Criterion and Discriminative Information Criterion to find the best models. Further improved sentence predictions by implementing a Uniform Cost Search which constructed a guess by scoring a search path using both the probability for a given word provided by the HMM model as well as the trigram Language Model score of the path.
Stacks Image 641
Machine Translation Model | view on GitHub
Built a deep neural network with embeddings and an encoder-decoder architecture that functioned as part of an end-to-end machine translation pipeline. The model accepted English sentences as input and returned French sentences as output.
Stacks Image 649
Facial Keypoint Recognition System | view on GitHub
Built a facial keypoint recognition system by combining computer vision techniques and deep learning. I extended the project to use the detected facial keypoints to overlay a “sunglasses” filter on a real-time camera video stream.
Stacks Image 657
DNN Acoustic Model | view on GitHub
Experimented with and designed a number of deep neural networks to function as the acoustic model step of the ASR pipeline. The model accepts raw audio as input and outputs a transcription of the audio.
Stacks Image 665

Check out Udacity's AI Engineer Nanodegree to learn more about the program.