Collaborated in product design reviews and helped craft requirement documents for the development team.Led a team of remote engineers in the development of a new social media mobile application which allowed communities of faith to form and communicate in a safe environment.  The app was built using React Native/Typescript on the frontend which communicated with a GraphQL API developed in Elixir/Phoenix/Absinthe.  In addition to the GraphQL API the mobile app also used several 3rd party APIs for authentication (Auth0), notifications (Getstream) and realtime audio session (Agora).Setup and maintained a robust AWS infrastructure which includes: AWS account per environment, VPC per environment with multiple subnet tiers, RDS database setup in each environment, EKS Kubernetes cluster  setup in each environment, Route53 zone/entry setup in each environment, OpenVPN bastion host setup in each environment, CloudWatch monitoring/alerting/logging setup in each environment, automated SSL/TLS cert management with AWS cert manager, secrets management with AWS secrets manager.  All of this is orchestrated with Terraform.Lead a second team of remote engineers in the development of a second iteration of the social media application.  The new app was designed as a short video sharing app where the primary intent is to express the things you are grateful for and to give gratitude to other users for the posts they have shared.  This app was also built with React Native/Typescript on the frontend and using Supabase on the backend.  This was an explicit architectural decision that traded backend flexibility with an accelerated development timeline on the frontend.

Designed and implemented a chatbot using the Rasa framework. The chatbot allows users to perform “conversational planning” over text messaging in order to add tasks, schedule reminders, etc. The chatbot can be interacted with using either SMS or Facebook Messenger channels. The chatbot is built with machine learning and training data in order to build a model which can then be used for intent classification, entity extraction and also assisting with dialog state transitions.Setup and maintained the system infrastructure which was primarily a GCP/GKE Kubernetes cluster along with some addition GCP cloud services (DNS, Storage, Service Accounts). Terraform was used to setup the GKE cluster and other cloud services. Helm was used to setup and manage the services running within the GKE cluster including: Rasa, Duckling, Nginx-Ingress, Cert-Manager, External-Dns, Postgres, Kafka, ElasticSearch & Kibana. The goal of the architecture was to run as much as possible in Kubernetes and to be as cloud provider agnostic as possible.Designed and implemented marketing website which allows users to request an invite to the system as well as to signup and setup a basic profile in the system.

Team lead on core engineering group which is responsible for collecting, processing and storing profile and content data from “influencers” across a variety of social media platforms. In this role, I lead a team of 4 engineers and helped facilitate and coordinate work on technical roadmap items that improved the teams operational efficiency.Designed and implemented reporting service API which manages metadata associated with Influencer Market Benchmark (IMB) reports. The service is implemented in Elixir and uses Phoenix/Absinthe to provide a GraphQL API and Ecto/Postgres to store/access the reports metadata. The Traackr UI application uses the services GraphQL API to access the report metadata information. These reports are defined and used by customers to gain insight into how their brands compare against others in the same market/industry.Migrated core applications to run as resources in Kubernetes within an AWS VPC environment. Part of this work involved redesigning the CI/CD process for all core applications which resulted in a consistent and flexible way for applications to define their own build/deploy pipelines.

Designed and implemented Yieldbot's next generation ad serving platform by porting the existing implementation from Node.js to Elixir. This resulted in a 16:1 reduction of memory usage and 3:1 reduction of EC2 instances required for the same load. The ad server was designed as a set of independent OTP applications. One of these OTP applications provided an HTTP API using the Phoenix framework and was designed as a set of reusable Plug middleware modules.Increased scalability and availability of the ad server's transient state caching layer by moving from Redis to ScyllaDB. The transient state caching layer is used when handling ad serving requests to correlate bid requests with downstream events such as creative rendering, click requests, etc. Moving from Redis to ScyllaDB allowed us to scale the cache databases horizontally and also eliminated Redis as a single point of failure.Established system of record database for advertiser/publisher configuration data using Datomic. Datomic implicitly tracks all changes to the database and allowed us to see the history and state of the database at any point in time. Established system of record metrics database for ad serving revenue generating events using ElasticSearch and Kibana. ElasticSearch enables the running of queries/aggregations on virtually any dimension field defined within the event records. Kibana allows business stakeholders to perform ad-hoc queries and create dashboards to gain realtime insights into various publisher/campaign performance metrics.Established application runtime environment with all applications running in Docker containers within a Mesos cluster. The Singularity scheduler framework is used to manage long running services, one off tasks and periodic scheduled jobs all within the same cluster. A load balancing layer using AWS ELBs, HAProxy and the Baragon orchestration framework provides seamless zero downtime deploys for all applications running in the cluster.

Lead Software Engineer working on Kayak's hotel search backend systems. This included creating/updating search spiders that would query various hotel search providers for availability when requests were made to the Kayak consumer UI. There were many different APIs that the hotel search providers needed to interact with including XML, JSON and also some HTML screen scraping. Because of this, the search providers were implemented as adapters that normalized their search results to a common data format.

Lead Engineer working on the continued development of the Mindreef SOAPscope SOA testing application. In June 2008, Mindreef was acquired by Progress Software.

Lead engineer on SessionManage J2EE based network management system.

Lead engineer on J2EE based network management system for DN 7000 Series Multiservice Switch platform (formerly from WaveSmith Networks).