Starry is a fixed-wireless internet service provider that develops and deploys proprietary wireless hardware.

I was promoted to manage a team responsible for release engineering, internal tooling and infrastructure. We supported firmware development, validation and deployment teams. I was responsible for architecting solutions, allocating work, training, and meeting quarterly goals.

Developing firmware images can be a very complicated and user-unfriendly process, especially when you are trying to integrate 5+ products, with 100+ repositories, and a 2+ hour build. An enormous amount of software and infrastructure are required to support teams who are developing, validating, manufacturing and deploying firmware at scale.

· Automated software integration from first commit, to final release. Our prior process had required timely, manual intervention from software team leaders. Manual intervention was replaced by a bot which responded to events from Github and our CI tools. The system automatically created its own code changes, pushed commits, performed reviews, and merged changes. Written in Python, Rust, Groovy and Typescript.

· Developed system for configuring newly manufactured antennas by non-technical users in our manufacturing facility. The system utilized a Rust web application running on a Raspberry Pi which would make a network bridge to out antennas, then configure them using Protobufs. The user was able to control all of this with a web application instead of a command-line interface.

· Launched and managed a Kubernetes cluster that support several teams on Google Cloud Platform. Developed Helm charts to deploy a variety of new services like continuous integration, a remote firmware update service for Starry devices, documentation, nightly firmware builds, and various services for our QA team.

· Developed tools to support firmware release engineering in modern Python3 and the Git API.

· Firmware projects such as language bindings and system utilities.

Liberty Mutual is replacing their proprietary call-center with a solution built in-house using modern tech. I worked on call-center features, including Natural Language Understanding components of the service.

Java development using Spring Boot, DynamoDB and API.AI.

DataRobot is a machine learning platform that allows data scientists to build and deploy accurate predictive models.

Predictions Squad (2015 to Present)

· Helped develop accurate and performant, high-throughput, low-latency prediction API using Python, Pandas, uWSGI, Redis, Nginx and MongoDB.

· Developed client application for high-performance predictions in Python. Doubled the scoring speed, fixed memory leaks, and developed it work across platforms (Linux, OSX and Windows) as a single-file-executable with no dependency on the Python interpreter. Utilized Requests lib, multiprocessing, data pipelining, multi-threading and network compression. Used TravisCI and Appveyor for Linux, Windows and OSX testing and artifact building.

· Helped develop test library that tested prediction consistency across all products. Used PyTest framework and Pandas dataframes for analyzing data. Ensured Python2 and Python3 compatibility

· Wrote automated performance benchmark tooks for prediction API using Python, JenkinsCI and Pandas.

DevOps team & Enterprise Release team (2014 to 2015)

· Helped develop a cloud-based SaaS distributed application into a Dockerized on-premise installation for enterprise customers. The cloud-based deploy was managed with Ansible, and used Amazon S3 for storage. I helped develop the on-premise install which utilized Ansible, Make and Docker for the deploy. I replaced the dependency on Amazon S3 by utilizing GlusterFS's API from directly within the application, rather than mounting a FUSE volume into the container.

· Fully provisioned, deployed and secured serveral Hadoop clusters with a mixuture of Cloudera Hadoop and Hortenworks Hadoop for product testing and Business Intelligence. Then I created an automated Cloudera Hadoop deploy process using their Python API for release testing automation.

· Build infrustructure for cloud-based SaaS distributed application using Amazon EC2, ELB, Route53, Auto-Scale groups, Terraform, Cloudformation and Deis.