Steve U. Email & Phone Number

San Francisco, California, United States

High-speed geometric algorithms, concurrent data structures. Very fast sorting.

San Francisco, California, United States

High-speed financial calculations and data structures.

San Francisco, California, United States

I implemented and helped to design cool high-speed mixed first and second-price auction technology for marketplaces and their vendors, including budget pacing, autobidding, quality scoring, and dynamic goal setting for ROAS.

Managed three data engineers combining data sources from third party products (Salesforce, Hubspot, Stripe, Chargebee), and source types (Segment, Aurora) into a queryable warehouse (Redshift) on top of a shifting landscape of active schema changes (Ruby on Rails, Salesforce) and changing business priorities.

Built a predictive model to preorder sales leads that resulted in a +74% conversion rate over the previous unordered process. Built models for rank ordering active accounts and sales opportunities in progress, and evaluating the health of individual accounts. Built models for Event (product) success that outperformed NPS (surveys) for predicting business.

San Francisco Bay Area

Built low-level lock-free and partially locking math libraries for error-correcting, error-detecting, and erasure-correcting codes, in particular Reed Solomon and XOR-only codes with a wide range of symbol sizes. The product is a thin, transparent layer on top of UDP that gives drop resilience. Other applications are possible. This was mostly to learn.

San Francisco Bay Area

Designed and taught basic data science, math and computer science topics and seminars in the asia-pacific region, europe, and the united states, both in-house and for the public.Built a maritime-safety classifier that mixed time series from disparate sources with different time synchronization schedules indexed on the same key to build a fixed-timeline.

San Francisco

Applied ML & stats techniques to fraud ranking (i.e. which customers were most likely to commit fraud in the future).The backend implementation was similar to a conjunctive normal form filter, so I found all of the separating planes for the logistic regression model (i.e. cutoff combinations that crossed the decision boundary) and turned it into a series.

Palo Alto, CA

I found a way to cut 30% of the expense of google adwords marketing without reducing customer lifetime value or volume of new customers. This was altered for other marketing channels and products as well, including facebook, sales via the mobile app, etc.I found a business process change in the sales department that pointed to a large gain possible from.

Buffalo, NY

Research in error-correcting codes, fault detection and provable statements about data storage. Defended Dissertation on 12/13/2012. Awarded in Feb, 2013.Research Assistant, 2007-2012Teaching Assistant, Fall 2009, Intro to Computational ComplexitySummer Research Internships, 2008 and 2011

Palo Alto, CA

Intern in the SIMPL storage research group at HP Labs. Worked on a complexity project and a systems project, both related to NoSQL storage.

Santa Clara, CA

Wrote a sampling data deduplicator in C for estimating compresibility. Also wrote code in C using FFTs of symbol entropy for the same purpose.

Phd, Computer Science And Engineering

Activities and Societies: Go ClubAlgorithms, Theory of Error-correcting codes, Complexity Theory.

Bs, Ba, Mathematics, Art

Activities and Societies: Water Polo, SwimmingMathematics degree with a concentration in Number Theory and Algebra. Art degree with a.