Geni — one shared family tree for humanity

Geni was one of the earliest companies to fully embrace Ruby on Rails. We were not simply building applications with Rails—we were helping shape the ecosystem.

The engineering team contributed open-source libraries, presented at conferences, shared practices with the community, and demonstrated that Rails could power a large-scale consumer internet platform.

At a time when many people questioned whether Rails could scale, we were proving it in production every day.

One of the biggest lessons I learned at Geni was that platforms outlive features. Instead of repeatedly building isolated functionality, I found myself designing systems that entire teams and products could build upon.

That pattern has followed me throughout my career. I naturally gravitate toward building platforms that enable other people to ship.

The original Newsfeed architecture had been created by early Geni engineer Adam Pisoni. He built its Ruby map/reduce core, called it Skynet, and later presented the system at RailsConf.

When David Sacks came up with the idea that became Yammer, Adam transitioned to the new product and eventually became its CTO. I took ownership of Geni’s Newsfeed platform and evolved it into the system that continues to power the product.

At first glance, a newsfeed appears simple. In practice, every action can generate a story, and every story must account for actors, targets, embedded objects, hyperlinks, permissions, formatting, localization, aggregation, caching, and delivery.

The challenge was not displaying activity. The challenge was describing it. The Newsfeed became a storytelling engine rather than a list of events.

Story aggregation was especially important. Instead of flooding people with repetitive updates, the engine grouped related activities into concise narratives while preserving access to the underlying details.

The Newsfeed exposed a problem that traditional localization systems could not solve. Normal translation frameworks assumed software consisted of plain strings. Our stories contained hyperlinks, embedded people and albums, grammatical gender, pluralization, variables, and nested formatting.

I began designing what became Translation Markup Language, or TML, and Justin Balthrop and I built the tr8n localization platform around it.

Instead of translating text, we translated structured stories. The framework understood links, variables, grammatical rules, gender, plural forms, and language-specific word order while allowing translators to produce natural translations.

Justin was also the brains behind Geni’s Tree Positioning System, or TPS. He created custom graph algorithms capable of lightning-fast searches across the global family tree—an unusually difficult problem at that scale.

Toward the end of our time there, we explored combining TPS and tr8n into a relationship translation system. Given the path coordinates between any two people, it could describe their relationship naturally in any supported language. I am not sure the experiment ever launched after we left, but it connected two of Geni’s most sophisticated systems in a fascinating way.

TML and tr8n attracted significant attention after David Heinemeier Hansson shared the framework publicly and ultimately inspired me to spin the technology into Translation Exchange.

Localization became a living platform rather than a one-time translation project. Instead of relying entirely on professional translators, we enabled the Geni community to translate the product collaboratively.

As Geni evolved, the community continuously improved its translations alongside it. Users translated the product into more than seventy languages.

I designed and built Geni’s public developer platform before modern developer portals became commonplace.

It gave external developers a self-service path to safely extend Geni while preserving platform consistency, authentication boundaries, and API versioning.

Subscription billing presented another foundational challenge. Stripe did not exist, so recurring billing, lifecycle management, free trials, payment recovery, and provider differences had to be engineered directly.

I architected Geni’s subscription platform around Authorize.net, PayPal, and ActiveMerchant. Many capabilities developers now receive from managed billing platforms had to be built and operated from scratch.

Communication became another reusable platform. I designed Geni’s email infrastructure to support high-volume, personalized communication across product and account lifecycles.

Localization and template management were part of the infrastructure itself, allowing product teams to communicate consistently without rebuilding delivery logic for every feature.

Geni fundamentally shaped the way I think about software. It taught me that the most valuable engineering work is often not an individual feature—it is a platform that enables an ecosystem.

Many ideas I continued building through Translation Exchange, developer platforms, healthcare AI infrastructure, workflow automation, multi-agent systems, and AI orchestration can be traced back to lessons learned while building Geni.

It was not simply another startup. It became the foundation of how I think about software architecture.