Erik Brinkman

• Developed SOTA calibration-based recommendation based fairness metrics, and applied them to internal systems.
• Researched, created, and deployed state-of-the-art large-scale sparse user-author-content models across Instagram, Integrity, Content-Understanding, and more.
• Developed transformer-based two-tower user-content models for cold-start unconnected content discovery that achieved half of our topline goals.
• Regularly conducted offline and online model analysis using SQL, python, and other necessary tools.
• Led ongoing project to learn hierarchical structure on top of existing user-content modeling.
• Advised two PhD interns on successful internship projects, resulting in accepted return offers.

• Studied auction settings where having access to public information is bad for all agents (AAMAS 2014)
• Demonstrated surprising efficiency of continuous double auctions in the presence of strategic agents (Submitted TEAC)
• Investigated how to design a modern frequent call market using empirical mechanism design (EC)
• Analyzing "Flash Crash", metrics that indicate susceptibility, and policies that mitigate risk in the presence of strategic traders

• Trained content embedding model of job openings using caffe2
• Doubled click-through rate for suggested jobs surface

• Modified auto-bidder to increase advertiser return on investiment
• Ran live experiments to determine modification effects and optimal parameters

• Developed search ranking optimization based on click-through data
• Built profile photo categorizer for demographics and search ranking

• TA'd introductory Algorithms and Data Structures course

• Implemented an efficient topic modeling algorithm (PLSA) in MapReduce
• Prototyped real time entity resolving and disambiguation system for news corpora
• Designed a database that supports general queries on geographic descriptions
• Adapted aggregate movement model between data modalities

• Studied the Cramér-Rao Lower Bound (CRLB), and it's implications on polar coordinate tracking with both continuous and discrete noise
• Built a Matlab simulation to investigate the relation between Extended Kalman Filter tracking error and CRLB for important target trajectories

Thesis: Understanding Financial Market Behavior through Empirical Game-Theoretic Analysis

Honors: STIET Summer Research Grant Recipient · Eta Kappa Nu · Microsoft College Puzzle Challenge Finalist (2015)

Selected Courses: Machine Learning · Online Learning · Reinforcement Learning · Social Networks

Honors: Gregory Sullivan Award for Professional Achievement · Tau Beta Pi · Theta Xi · Dean's List

Selected Courses: Algorithms for Nonlinear Optimization · Nonlinear Dynamic Systems