Tech Lead Manager for Feed Ranking Team

Tech Lead for Feed team

Teaching assistant for C programming (CS 11), Intro to Complexity Theory (CS 21), Algorithms (CS 38), Machine Learning (CS 156a), Intro to Systems (CS 24). Held weekly office hours, recitations, reviews. Graded problem sets. Wrote TA notes and solution sets.

Implemented a model of mouse behavior using imitation learning.Built and trained adversarial recurrent neural networks in PyTorch for general imitationlearning tasks, and applied model to resident-intruder mouse dataset.

Studied Bayesian hierarchical methods applied to semi-supervised learning under the Caltech Summer Undergraduate Research Fellowship. Designed spectral clustering algorithms with Markov Chain Monte Carlo sampling, specifically with modified Metropolis-within-Gibbs algorithms. These hierarchical algorithms sampled not only the classification, but also the distribution of hyperparameters that govern the shape of the prior distribution of the classification. Designed experiments to advance understanding about the behavior of these hyperparameters and the best subsets of hyperparameters to learn.Implemented these algorithms in MATLAB and measured their convergence rate and classification accuracy on real datasets such as MNIST digits and U.S. House of Representatives voting records, and synthetic datasets such as the two moons dataset. Improved the convergence rate of the multiclass clustering algorithm by introducing an adaptive step size, which uses the observed acceptance probability to increase or decrease the steps. Finally, compared hierarchical and nonhierarchical algorithms and observed advantages of the hierarchical approach.

Built a distributed system that subscribes to multiple market data sources, aggregates, and reports trades. Implemented interning of symbols and routes and replay mechanism to catch clients up on missed trades.Built a proof of concept of an improved risk management system for the options desk. Wrote output to PostgreSQL tables, improving usability by allowing for custom queries, and improving scalability by allowing for techniques such as sharding the database.

Worked as a machine learning engineer on the content distribution team. Engineered a reinforcement learning system from end-to-end. Built a system for notification sending and designed the system to be easily adaptable to different problem domains, paving the way for more reinforcement learning solutions. Improved ranking for Feed, Ask to Answer features by building new models and analyzing A/B test results.

Instructor for the MathCounts math olympiad class at A* Summer Math Camp.Taught classes in Algebra, Geometry, Number Theory and Combinatorics.Improved handouts, wrote exercises and solutions and updated lessons.

Developer for a SysML diagramming plugin for MagicDraw that allowed systems engineers to design flight systems and generate the appropriate XML and C++ representations.Wrote the plugin in Java, using the MagicDraw API to check for errors and inconsistencies in user-created models. Learned to write JUnit tests to aid development. Learned to use Apache Velocity to generate xml documents.