Joe Bylund's Resume

Previous Employment

Altana - ai powered solutions for shipping/logistics/customs/government - supply chain intelligence
Spring 2022 - Summer 2023

Business network discovery and traversal - built tools to construct business networks from graph of world wide shipment data.
Search microservice - Built search microservice for companies, facilities and transactions in the Altana atlas using FastAPI, Pydantic, ArangoDB.
Arango client library - built http client for ArangoDB supporting synchronous and asynchronous requests used in search mircoservice.
Spark client library - shimmed three different client libraries (pyspark, pyhive, databricks-sql-connector) to pep-249 interface, supports creation of business networks.
Geocoding client - wrote geocoding client library which handles address pre-processing and http requests to a pelias service. Built tooling to measure geocoding accuracy over time.
Geocoding pipeline - built continuous address geocoding pipeline and geocoded the ~400 million addresses in the Altana atlas.

Kensho - ai solutions for the financial industry
Summer 2018 - Spring 2022

Designed and implemented multi-step document processing pipeline using multiple in-house ML services.
Designed and built speech-to-text alignment pipeline using SQS & gentle forced aligner.
Implemented a number of checks in github hooks (pylint, flake8, mypy...), improving the developer experience.
Migrated fuzzy company identification service to kubernetes and optimized performance.

Moat - Advertising analytics for advertisers and publishers
Fall 2013 - Summer 2018

Designed and implemented distributed, fault-tolerant ETL, reducing cost by an order of magnitude, increasing reliability, and reducing processing time from ~10 hours to ~1 hour, making data available to clients far earlier in the day (using python, SQS, Redis, PostgreSQL).
Numerous data-driven API improvements which lead to 3-4x improvement in API latency as well as maximum request size (PHP, CakePHP).
Contributed improvements to ORM (CakePHP) and PHP postgres driver in order to reduce the number of queries necessary to render a page by 5x (decreasing page load time by ~3x) (CakePHP, c).
Standardized deployment framework used to deploy thousands of servers of ~30 different roles to AWS (AWS, EC2, boto3).
Migrated primary non-statistical database (users, accounts, metadata) from MySQL to PostgreSQL, improving uptime & flexibility (MySQL, PostgreSQL, foreign data wrappers).
Migrated primary statistical database (500 million rows/day) from non-first normal form to first normal form schema, improving query latency, reducing storage demands, and increasing throughput.
Architected and implemented sophisticated message routing system which is responsible for moving ~40 billion events per day from pixel servers to real-time processing servers while balancing CPU and memory constraints (c++).
Architected and prototyped massively parallel decentralized data lake using AWS lambda and S3 for cost-effective storage and low-latency and cost-effective queries (AWS lambda, python, PostgreSQL)

2007 - 2013

Led development and maintenance of Protein Local Optimization Program (PLOP) project, a molecular mechanics library for protein structure prediction (fortran).
Designed and implemented the computational mutation scanning module of PLOP.
Redesigned build system to automatically determine dependencies and take advantage of parallel compilation, reducing build time from ~30 minutes to ~3 minutes and accelerating development.
Created a small molecule database representing 95%+ of small molecules in the Protein Data Bank, extending PLOP from a protein-only program to a general molecular mechanics toolkit.
Designed and implemented a Perl based automated regression testing framework, which accelerated development while minimizing bugs and regressions.
Created a project wiki, combining scattered documentation and completing missing documentation.

2003 - 2007