Patrick Cockwell

I helped design, vet, and implement a robust data model for use with data from various clients in the airline industry.

Each client has unique internal representations of the data, which are transmitted to the organization data stores on varying schedules and frequencies using an Airflow ETL platform. The data is then processed and stored into the canonical data model (CDM) such that the machine learning (ML) models generated by the data science team can operate on the data.

This CDM is used across all clients and requires data idempotent and custom transformations for each client.

I remodeled the entirety of the company infrastructure on Google Cloud platform to provide increased security, faster deployment of new infrastructure, data and infrastructure isolation, and fine-grained permissions. This was done using Terraform to write multiple reusable modules that were combined into modules of reusable segments of infrastructure, and finally combined into a full infrastructure deployments.

I helped model and make modifications to an existing company infrastructure in Amazon Web Services to follow best practices for highly available redundant systems to provably be able to sustain the infrastructure, applications, and data in the event of an infrastructure outage.

This involved data replication across AWS availability zones (AZ), and changes to infrastructure to shift the load across multiple AZs. Once the infrastructure changes and data replication were made, a full availability zone failure was simulated by terminating and stopping all resources in a single AZ. We provably showed that the system remained resilient, and no data loss occurred.

I built a visualization tool using PHP, HTML, and JavaScript for timeseries data stored in a Graphite data store.

This tool enabled the user to select multiple types of graphing capabilities including data breakdowns across metric subcategories and period lookbacks (e.g., week over week comparison).

Users were authenticated using one of the multiple methods (OAuth being the primary means). They could also create and save dashboards for use across the company, add external links, and drill down into an individual graph or dataset.

This tool was used to help diagnose and notify Apple of App Store payments outage in June 2012 prior to their knowledge outage. The Orion Analytics Platform was developed at Funzio, which was purchased by Gree International in May 2012, and was open sourced in Aug/Sept 2012.