Shannon Carey Email & Phone Number

Chicago, Il, Us

Seattle, Wa, Us

Seattle, Wa, Us

Joined AI Labs

Seattle, Wa, Us

On rotation with Traffic Engineering, productionizing ML model for detecting traffic from bots with AWS SageMaker, and providing tech ideas/strategy for path forward from current human-driven flat list of yes/no rules toward a more flexible system that can integrate human curated rules with enriched data and input from Machine Learning etc., and combine it using Bayes-inspired representations of certainty, to produce more sophisticated traffic treatment and characterization.

Seattle, Wa, Us

"You're a manager now??"Yes, and I still code, just not as copiously as before.

Seattle, Wa, Us

Orbitz was acquired by Expedia.Main projects:- A platform to allow teams to write & deploy lambda functions which execute against events from Kafka topics, and produce output in a standard format.- A platform to build and use user profile data by allowing teams to create and deploy event-driven jobs which eagerly pre-calculate and store various attribute information on a per-user basis, for the purpose of providing fast lookup of those attributes during the website request-response cycle in order to customize the website experience.

As technical lead, created a system to analyze A/B test performance by gathering click-stream data into AWS Kinesis, processing it in a distributed & intra-machine parallel manner with ReactiveX (RxJava), inserting it into RDS and Redshift in as optimal a structure as possible (considering data partitioning & querying patterns), and performing Bayesian analysis on the result. Results were presented in a Grails UI with a React.js replacement in the works. My team performed all dev-ops work for the system and made heavy use of CloudFormation and ElasticBeanstalk for provisioning infrastructure and performing deployments. The click-stream collector was deployed in multiple AWS regions and made use of latency-based DNS resolution for better performance in Europe & the eastern hemisphere.The main challenge on the project was the need to join each new data point to the user's activity from the entire history in the database and the need to perform a count of distinct values in high-cardinality collections. In addition, we were required to design a system that could respond to various user-controlled queries in real time (no batch pre-processing). We investigated HyperLogLog as a means to vastly reduce the amount of data needed, but the suggestion was ultimately rejected over concerns about loss of accuracy.Key takeaways:- If you're trying to accomplish 0% data loss on data transferred from browsers over HTTP to AWS ELB, that's probably an unrealistic goal. The internet is not reliable.- Avoid predicting the future. Avoid complexity that is retained because you think it will eventually be useful (YAGNI). This can be difficult when you're unable to accurately test the performance of the system until the data is at-scale, but your team will appreciate retaining their sanity until they need to solve the problem when it actually occurs.

Chicago, Il, Us

Full-stack lead engineer for Data Garden (www.datagardenco.com), a web-based platform that gathers, analyzes, and provides insight into sales data for natural and organic manufacturers in order to help them make better choices for their business.Led development of a major component of Satori, SPINS' new business intelligence platform.

Worked on a small, startup-style team to create a new software platform for Wrapport's hyper-local news sites in the USA and abroad. Full-stack development made use of Ruby on Rails for both RESTful API and separate frontend component, MongoDB, provisioning with Chef, and cloud hosting on AWS, Helped integrate a dependency graph on Neo4j to enable accurate cache invalidation.Enjoyed working with (and sometimes contributing to) a variety of open-source software in the Github and Rubygems ecosystem.

Denver, Colorado, Us

Writing GIS software for the world's largest producer of navigational charts and data for airplanes.The key aspect of this job was dealing with bitemporal data and supporting it in various aspects including: data modeling, data management, programming API design, and distributed scaleability of bitemporal data storage and computation.On the server side, worked on implementation of map-reduce algorithms and approaches for spatial data indexing in in Oracle Coherence. On the client side, performed prototyping, technology choice, and subsequent implementation for the spatial and bitemporal data visualization and editing GUI tool.

Writing GIS software for the world's largest producer of navigational data for automobiles.Besides rooting out tricky Swing event listener and 2-d geometry bugs and implementing new features, I was proud to have been a key part of the creation of a new visualization mode for the GIS editing software. The project relied on an understanding of trigonometry and linear algebra, and came with plenty of algorithm design opportunities, UI design opportunities, and challenges for providing a rich representation of a detailed but difficult data set.

Normal, Il, Us

I led a team of student software developers for a year after graduating. We developed web apps that increased productivity for campus employees and offered enhanced services to students. In addition, the group helped introduce students to the experience of designing, developing, and shipping software from the ground up as a team.