Aleksei Sholokhov Email & Phone Number

Seattle, Wa, Us

Seattle, Wa, Us

South San Francisco, California, Us

Seattle, Wa, Us

* Created an algorithm that extracts interpretable models and learns physics laws from trained neural networks. Decreased the amount of data needed to model noisy dynamical systems by a factor of 10.* Accelerated training of deep neural networks with higher-order methods. Implemented it as TensorFlow and jax modules. Achieved state-of-the-art performance for selected image recognition tasks.

Seattle, Wa, Us

* Learned OpenMP, MPI, and CUDA by working as a teacher assistant for graduate-level High-Performance Scientific Computing classes for two consecutive iterations of the course. Developed expertise in MATLAB by holding office hours for Scientific Computing classes for 1 year in total.* Developed gspack: python-autograder to accelerate grading of coding assignments. This software has become a critical infrastructure for the department. It has been used in 5 scientific computing classes for over 3000 students in Department of Applied Mathematics.

South San Francisco, California, Us

Designed and implemented a calibration pipeline for large deep learning models using flyte framework. Improved the target metrics by 300%. Enabled the team to offer their products to a broader range of downstream consumers.Transformed my team's vision into a project proposal. Communicated extensively with my leadership to ensure meeting the company's needs. Presented 3 times at team-wide meetings. Drove the project to completion in 3 months.

Cambridge, Ma, Us

Created a new deep learning algorithm for predicting behavior of physical phenomena using an embedded device. Implemented it using pytorch and tensorflow. improved the target metrics by 250%. Drove 1 paper from proposal to completion in 3 months and contributed, as a second author, to 1 additional paper.

Seattle, Washington, Us

As a part of Math Sciences Team:* Invented new statistical modeling tools that extract meaningful features for machine learning models. Implemented it as a python package "pysr3" that is fully compatible to scikit-learn. Achieved 30-fold speed-up upon deployment to the institute's pipelines.* Developed IHME Projections, a statistical model that projects cases and deaths from COVID-19 globally, in cross-functional collaboration with a team of 130 researchers. This tool helped the decision makers (including local, state, and national governments) to manage resources and plan ahead during the pandemic.

Los Alamos, Nm, Us

* Developed a reinforcement learning framework for controlling ensembles of thermostatic devices. This project resulted in my undergraduate thesis.

Doctor Of Philosophy - Phd, Applied Mathematics

Master Of Science - Ms, Applied Mathematics