| Contact: evulfson@gmail.com Remote (US-based) github.com/yegeniy yegeniy.github.io 781.632.3780 |
Executive Summary:
calm + conscientious + creative = ships productA Principal-level Engineer with over a decade of experience in machine learning, data engineering, and software engineering. Skilled in Python, Scala, Java, and AWS technologies. Proven expertise in making cross-organizational impact, building lasting data and backend systems, iterating on the full machine learning flywheel, and guiding teams through significant transitions. |
|---|
| A Silver Revolver | I strive for a tenfold improvement in my teams' impact through a pragmatic combination of techniques from software engineering, organizational psychology, and personal experience. |
|---|
"No Silver Bullet" is still true, so 10x software teams by loading "A Silver Revolver" with these bronze bullets:
pithily: agile, async, git, nbdev, PRs, Sublime Text, CLI, testing, and a lot of Phind
Eugene Wolfson - (Remote) United States of America
I’ve been immersed in Machine Learning and Data over at Meetup for over half a decade. Mostly Python and Scala and AWS. Prior to that, I spent half a decade at MLB Advanced Media doing all sorts of backend software engineering in the live streaming media space. This is the company that became Disney+. Mostly Java and AWS with a good chunk of Python at the end as I transitioned into data. And before that I spent a couple of years at an e-commerce startup called OpenSky. Mostly doing backend engineering in Java and PHP and deep diving into how startups work. My higher education concluded with an Java software internship at a beautiful non-profit called Trace Foundation. I completed a Masters of Arts in Computer Science and a Bachelors of Science in Physics, Mathematics, and Computer Science over the course of half a decade at Brandeis.
2018 - 2024
Meetup is about making Babies, Bands, and Businesses.
I worked on the data team, making team contributions to the Machine Learning Engineering, Data Engineering, and Data Science aspects of the team. As well as providing ad-hoc assistance to critical infrastructure and backend problems of the company at large.
It's been over half of a glorious decade of contributing to Meetup's data team. Everything from Machine Learning Engineering, to Data Engineering, to Data Science. A lot of Spark, AWS, Jupyter Notebooks, Python, Scala, experimentation, SQL analysis, and recommender systems. With very little focus on the Business Intelligence (Looker) reporting side though. My career focus here has always been on training ever-better machine learning models. But that's just the pearl inside the oyster at the bottom of the data lake. Prior to being able to model anything in a company as small as Meetup, we need to build, maintain, and operate all the systems that get us the data. And then do the same for all the systems that serve out the recommendations at scale. For context, the data team's systems are mostly devoted to generating recommendations and business intelligence. We also maintain a simple experiment assignment service called variant that all clients invoke for running split tests across the company. But the vast majority of the data team's endpoints are thin layers for serving out our recommendations. These are exposed as a collection of data APIs. They serve our recommendations out mostly over HTTP or in a few cases through SQS. The recommendations are generated from models, both trained/ personalized and not. The data for these recommendations and for our business intelligence comes from our S3-backed data lake. The lake is managed by an AWS MWAA (Airflow) instance which schedules a variety of ETL jobs. They extract data into S3 from various sources, such as MySQL and transforms it using Spark (EMR). Slicing and dicing it to ultimately land into either RedShift for reporting / analytics or into S3/DynamoDB/ElasticSearch for recommendations. Depending on the many use-cases our clients have.
2013 - 2018
Summary: Enterprise software engineering, with a focus on Java, multimedia, and conversational interfaces. This is the company that eventually became Disney+.
Platform for natural language interfaces (a.k.a. bots). A client and partner agnostic platform for creating natural language interfaces to MLBAM's existing services. Consists of the following parts: AWS Node.js and Python lambdas for the adapter layer for clients such as FaceBook Messenger, Slack, and Alexa; a Blue-Green deployment pipeline leveraging AWS ECS and docker containers, and Java 9 microservice containers built with Dagger2, Spark, Retrofit2, cukes-rest, junit/mockito, and lots of love and domain knowledge. These microservices are: an orchestration microservice to provide a consistent API for clients; a conversation understanding service integrating with wit.ai, and to hold conversational memory; a facade to integrate with the rest of BAM's backend service APIs; and a conversation analytics ledgering framework, backed by AWS' Firehose and S3 for anonymized analysis.
Key Encipherment System. A Java client that leverages AWS' DynamoDB and KMS services to provide a base level of protection for data at rest in an untrusted facility. In addition to providing a simple approach to storing encryption keys for our media, this project was also used as a real-world example for an internal talk underlining the value of designing multiple-region support into applications and services.
Distributed real-time monitoring framework. Some of the (snippet-sized) lessons learned on this project are available at http://yegeniy.github.io/archived-blog. Independent contributor to a distributed real-time monitoring system that collects data from various sources and adapts the aggregate into meaningful views for clients. Core Java libraries, Jersey (JAX-RS) with embedded Grizzly Server, Jackson for JSON, providing simple REST HTTP API to internal modules and external consumers. Long-lived TCP Socket communication and Protobuf messaging proof-of-concepts implemented. Features a sprinkling of compile-time code generation via Annotation Processing to reduce the burden of repetitive boilerplate code. The modules generally run under Tanuki’s Java Service Wrapper, but a branch for deploying to servlet containers exists.
M3U8 tooling (HLS) and Transport Stream parser (ISO 13818-1). Built in "low-level" style Java, mainly to become familiar with the specifications; however, it has turned out useful enough as a personal debugging tool to solve questions from co-workers occasionally.
Full-stack Java web application development for "multimedia" department. Previous role was full-stack Java web application development for MLBAM and its partners as a member of the "multimedia development" (10 people) and "multimedia framework" (5 people) teams. Work includes development, design, and upgrades of most of the team’s applications; automating testing at unit, functional, integration, and load levels; writing and organizing documentation; and above all working with, learning from, and mentoring co-workers (teammates, sysadmins, DBAs, and members of other teams). Participate in technical phone screens and interviews of candidates when given the opportunity.
Miscellaneous Software Development Life Cycle improvements. Spend downtime, when helpful, on finding and automating away sources of friction in the development lifecycle; always making sure to document information for others within and outside the team. Examples range from: necessary one-offs such as setting up CentOS user acceptance and test environments; to long-term iterative improvements such as improving the deployment workflow as new tools become available internally (various AWS services, Blue-Green deployments, Jenkins, internal frameworks, RunDeck); to quick returns-on-investment such as facilitating conversations among people who should talk or automating away persistence-layer operational overhead with simple bash scripts (CouchBase, MySQL); to occasional long-shots such as demonstrating Docker containers, chat rooms, git, or BDD.
2011 - 2013
Acquired a broad exposure to web development technologies, invaluable startup insights, and met some amazing people.
2010 - 2011
Raising a wonderful girl and boy! Early in my career, I implemented a personal GTD and Deep Work organization system variant to stay focused. Leveraged kimono labs and beautifulsoup4 to collect procyclingstats.com and espn.com data for academic research on teams. Learned bioinformatics by struggling through rosalind.info/problems; attended GenSpace’s Biohacker Boot Camp. One day, I'll finally pivot back into scientific programming. But in the meantime, I go on long walks, try to be handy around the house, and encourage friends who are learning programming and machine learning.
My recent reading list has been focused on Deep Learning and Generative AI literature and content. Basically it's all side projects, online courses, and the 2025 AI Engineer Reading List these days. But back in my software engineering career, I was keeping a proper reading list: Building Microservices, Smarter Faster Better and Clean Code. The previous year: finished Data and Goliath, The Inmates are Running the Asylum, The Power of Habit, Scala for the Impatient, How to Lie with Statistics, Liars and Outliers, Antifragile, Java Concurrency In Practice, Effective Java 2nd Edition, Magic of Thinking Big; and I stopped reading The House, in Which... (Russian) and The Well Grounded Java Developer. The previous year: finished Effective Java 2nd Edition, The Four Steps to the Epiphany, Principles (by Ray Dalio), and The Innovator’s Prescription; and stopped reading Clean Code, xUnit Test Patterns, Fundamentals of Biostatistics (by Rosner), and Pro Puppet.
Fluent in English and Russian, with basic proficiency in French. Taught high schoolers in an after school program called Russian School of Mathematics after work and on the weekends in 2017. Mentored children with Post Traumatic Stress Disorder as a Counselor for the Children of Sderot Camp in Israel in the Summer of 2009.
Way back in college, revived and became the president of Brandeis Physics Club from Fall 2007 until Spring 2010. Fostered interest in the sciences through local community outreach. Led three small groups in design and building of project for the Bernstein Festival of the Creative Arts. Wrote proposals to receive club funding from various on-campus sources.
Online Presence:
Teaching Assistant for Professor Mitch Cherniack Spring 2011
Student Researcher in Professor Seth Fraden’s Complex Fluids Group 2007 - 2010
Teaching Assistant for Professor Antonella DiLillo Fall 2008
Help Desk in Library Technology Services 2007
Biohacker Boot Camp. GenSpace. Brooklyn, NY 2015
Brandeis University Waltham, MA
M.A. in Computer Science - Spring 2011
B.S. in Physics with High Honors, Computer Science, and Mathematics - Spring 2010