About Jeff Sylvan · Jeffrey's Blog

Selected impact

Production ML leadership: led the first non-data-science engineering team at CarGurus to ship a production ML application with a full MLOps pipeline; the work was highlighted to investors and created a path for backend teams to deliver ML systems.
Enterprise platform adoption: established a continuous delivery pipeline at First Republic that reduced deployments from days to minutes, scaled across hundreds of engineers, and became the basis for training 50+ engineers, architects, and directors.
High-scale commerce rebuild: identified a tax-lookup scaling flaw threatening Cimpress EU sales, won cross-department buy-in, led the replacement, and handed off a platform that sustained thousands of queries per second without incident.
Founder/CTO execution: founded Apocoplay, led a 20-person team, and built large-scale multiplayer infrastructure that reduced server and development costs by 90% per game.

Currently: available for full-time Staff/Principal work, distributed systems, platform, and AI infrastructure. Boston area or remote (July 2026).

The short version

I'm Jeff Sylvan, a Staff/Principal software engineer focused on distributed systems, platform engineering, and AI infrastructure. Currently I am the founding engineer at CogniCluster LLC, an AI developer-tools startup, where I am architecting Black Skies and an AI-agent engineering workflow around Temporal, DAG-based planning, and agentic Playwright QA.

Previously I was a Principal Engineer at CarGurus; earlier work includes production ML infrastructure, delivery systems adopted by hundreds of engineers, commerce-platform rebuilds at global scale, and real-time multiplayer engines at two game studios, one of them mine. The receipts for the current work start at the build ledger.

Career record

Feb 2026 - present

CogniCluster LLC · Founding Engineer · AI developer-tools startup Architecting Black Skies, a persistent multiplayer space game designed for 10,000-player battles without time dilation, around strongly consistent H3 tile cells, a gossiped ownership mesh, Redis Streams, and Aurora Limitless. Building the work through Temporal-coordinated AI-agent workflows, DAG planning, agentic Playwright QA, red-team loops, 1,000+ tests, and chaos-suite coverage.

2023 - 2026

CarGurus · Principal Software Engineer Led the first non-data-science team to ship a production ML application with a full MLOps pipeline, highlighted to investors on an earnings call. ADRs and architecture reviews, monolith to services.

2018 - 2023

First Republic Bank · Senior Software Engineer, Principal scope Built the continuous delivery pipeline that took releases from days to minutes, drove adoption across hundreds of engineers, and trained 50+ engineers, architects, and directors on the practices behind it.

2017 - 2018

HubSpot · Senior Software Engineer Kafka-backed identity and user-data pipelines.

2015 - 2017

Cimpress (Vistaprint) · Lead Software Engineer Found the scaling flaw in core commerce infrastructure that threatened EU sales, won cross-department buy-in, led the rebuild, and handed a junior team a platform that ran incident-free at thousands of queries per second.

2012 - 2015

Apocoplay · Founder and CTO Founded the studio, led a team of 20, and built a server-side engine for large-scale multiplayer games that cut server and development costs by 90%, saving hundreds of thousands of dollars per title annually.

1999 - 2012

Earlier · ASC Games · Allscripts · Council for Logistics Research · PRTM · QScend Additional software engineering, infrastructure, logistics, and pre-release game work across the first decade of my career. Details on LinkedIn

Continual learner: University of Hartford, B.S. Computer Science · MIT certificates in Data Science and Blockchain · MIT Start Smart and Invest Smart · ongoing hands-on work in distributed systems, AI agents, platform engineering, load testing, and chaos testing. There is more depth on the resume and LinkedIn.

If you are evaluating me for a role: the resume is two pages, this blog is the work sample, and the fastest honest tour is the thesis followed by the ledger. Send a note and I will respond quickly.