Manish Kumar Singh Email & Phone Number

Mountain View, Ca, Us

Perception team

La Jolla, Ca, Us

- Online Learning for healthcare robotics: Worked with Prof. Kamalika Chaudhuri on an interesting application of the stochastic multi-armed bandit(MAB) model to healthcare robotics. The goal of this project is to model the sequential decision-making of one or more robotic agents interacting with People with Dementia(PwD) using the stochastic MAB model where each aspect of a robot’s interaction with a PwD is considered to be an arm of the MAB and the goal is to build online algorithms to make choices in order to maximize the expected reward. Studied a lot of theory of online learning algorithms, and worked on implementing a novel algorithm to transfer experiences in a multi-player scenario where different players can learn from each other. Work accepted for publication at Workshop on Real World Experiment Design and Active Learning, International Conference on Machine Learning (ICML), 2020- Dynamic relational inference in multi-agent systems: Currently working with Prof. Rose Yu on the problem of inferring relations in an unsupervised fashion from the trajectories of agents in a multi-agent system. Developing a Neural Relational Inference based model in PyTorch to infer relations and predict future trajectories of agents in a completely unsupervised fashion.

La Jolla, Ca, Us

CSE-105: Theory of Computation(Fall, 19)CSE-151A: Introduction to Machine Learning(Winter, 20 and Spring, 20)DSE-200: Python for Data Analysis(Fall,20)Responsibilities included preparing and grading assignments, holding office hours and discussion sections.

Redmond, Washington, Us

Research Fellow at Microsoft Research working in the Applied Machine Learning group- Built from scratch an embedding learning system using TensorFlow and C++ to implicitly factorize the HeteSim similarity matrix for node to node recommendation among heterogeneous entities for Microsoft Teams– Tested the embedding learning system on multiple benchmark datasets like MovieLens, Yelp and Douban Movie and reported improved performance on classification, clustering and ranking tasks– Implemented an explicit loss formulation of the HeteSim scoring model to train the embeddings using Tensorflow

Master Of Science - Ms, Computer Science

Bachelor’S Degree, Computer Science And Engineering