Volunteer computing is the process of harnessing processor cycles from multiple clients to contribute computing power towards a common external project. This method is most applicable to highly parallelizable, computationally intensive tasks. However, a major challenge in volunteer computing is handling client unreliability: clients can sporadically enter and exit the network, make computation errors, and require unbounded computation times. In this project we aim to design a volunteer computing schema that will support large-scale machine learning in terms of providing iterative stochastic parameter optimization. To accomplish this in a manner robust to error, we design a system for partitioning the the parameter optimization problem into smaller computational pieces and for combining the results.

I was first exposed to computer science during high school through computational biology research at Georgetown University and MIT. Since then, Ive enjoyed the opportunities to gain more experience through internships at Jane Street Capital, Evolv Technology, and Facebook.