Joseph Haaga

Grew the ML Platform team from 1 to 6 in first year (4 Senior Engineers + 1 Mid-level Engineer + Lead)

Built an ML Platform supporting 45 models across 4 environments, serving 1M predictions/day. (Terraform, EKS, Kubeflow, Istio, Knative, VPC, Oauth2-proxy, ArgoCD, Gitlab CI)

Introduced development environments for 20+ engineers providing elastic compute + GPU to support experimentation. (Kubeflow Notebooks, EKS, Istio)

Spearheaded the Interos Engineering Blog, establishing a content calendar and editorial board & authoring the first content.

Provided architectural/design oversight to breakout team working on building inner-sourced shared solutions. CTO called result "the most elegant solution to this problem I've ever seen"

Evaluated and implemented common capability for ETL jobs. (Prefect on EKS).

Designed & built `prefect-deployer`, shared dev tooling used in 455 deployments of 66 ETL jobs (and counting).

Advised on architecture for platform rebuild as part of a 7-person team chosen by CTO.

Mentored and supported engineers across the company, and published educational repos to upskill others on Kafka, fixtures & dependency inversion for cleaner unit tests.

Assisted in onboarding & orienting key engineering leadership hires (Director of DevOps, Director of Data Architecture, Chief Data Scientist)

Selected by VP of Machine Learning to spearhead ML Operations practice, resulting in promotion to Lead MLOps Engineer.

Supported efforts to achieve compliance with enterprise Cloud security mandates

Automated reporting for cloud infrastructure vulnerability scans

Led assessment & remediation of difficult cloud modernization efforts (Terraform, Ansible, boto3)

Led development of a graph analytics tool (React, Flask, Neo4j, Docker) for fraud pattern detection in financial data graphs.

Supporting research in applying novel deep learning techniques (RNN embeddings, Tensorflow, graphwave) to elicit graph embeddings.

Designed and implemented a image recognition application for environmental science use case.

Experimented with different convolutional neural network architectures to maximize F1 score.

Built custom sklearn Pipeline components to take advantage of OpenCV preprocessing techniques.

Developed web-based utility resulting in a 10x decrease in time spent labelling training data.

Led architecture and development of CMS-driven solutions for clients like Strayer & Global Automakers, using Drupal 8 and Pimcore 5.

Contributed to the architecture and development of marketing websites for brands like Strayer, Ligado, Politico, Merritt Clubs, JBG and EYA. Developed on Drupal 7, Pimcore 4/5, and Wordpress.

Drastically reduced development timelines by building internal tools to streamline internal processes (structured QA spreadsheet with Slackbot integration to close developer-designer feedback loop and closely track live performance/completeness metrics). Built using Google Apps Script, Google Sheets, and Slack Webhooks.

Delivered customer persona insights by applying data mining & visualization techniques to client CRM & transactional data. Performed analysis using Tableau and Anaconda toolkit (Seaborn, Pandas, NumPy).

Led development of a cross-platform mobile application to track a brand’s social media campaigns, using Apache Cordova and Adobe PhoneGap.

Developed and maintained digital presences for a variety of trade organizations, defense companies, and nonprofit clients. Developed on Wordpress and maintained on WPEngine managed hosting.

Developed automated spreadsheet parsers in Java 7.

Merged GIS shapefiles with political ad spend data to determine campaign efficacy in Designated Market Areas.

Prototyped IoT orthopedic rehabilitative device on Arduino hardware in a dialect of C.

Developed games to be played with the device to aid in gamification of proprioceptive rehabilitation exercises.

Implement wireless modules (XBee), LCD screens, & various sensors/actuators within I/O constraints of a development board.