Primer brings machine learning to new heights with impressive code base documentation and internal collaboration

Primer is a machine learning company that helps clients quickly find critical information, separating the signal from the noise in vast amounts of documentation. Its system can identify key insights and automatically write reports across topics like law, retail, finance, and national security. The San Francisco based start-up has raised $54.7M and counts government agencies and Fortune 50 companies among its clientele. The team has a diversity of experience, with some coming from computer science, but others joining Primer from the worlds of biology and economics. Like many fast-growing companies, they have run into challenges trying to document their code base and collaborate across different product teams.

The challenge(s): A fast-growing company needs their knowledge sharing to scale with them

William Du, a data scientist at Primer, found that he and his colleagues were wasting a lot of time on repetitive tasks, especially when it came to helping new hires. "We needed to build a repository of the questions we ask because a lot of them were being asked over and over again."

Primer also needed to find a good way to identify subject matter experts and to get the knowledge in their heads written down in a manner that was easily accessible and simple to search. "As you scaled the team, information transfer gets trickier,” says Amy Heineike, Primer’s VP Product Engineering and founding team member. “There are things you learn, or rather, some of you learn. The question becomes: how do you get that information to the rest of the team?"

Team members talked a lot through Slack, but this didn't work too well as a knowledge base. "It's notoriously difficult to trudge through your Slack history to find one question,” says Du. “GitHub didn't really make sense because we have individual products and they have their own repo so there's not really one place to put all your questions. And Confluence wasn’t working as a way to organize and share knowledge.”

Trevor Bossert, Primer’s DevOps Manager, felt there was an opportunity to put structures in place that would allow the company to scale successfully. "When I first took on the DevOps role, there was not a lot of tooling, no great automation, and a lot of decentralized decisions. We have a lot of different teams working for different clients, but there wasn't the ability to easily share things that they learned, errors that they ran into."

The solution: A knowledge base that could reach across department silos

Bossert decided to try out Stack Overflow for Teams. “I've been a very avid user of Stack Overflow, the public site, for a long time. When I saw that you were doing a private version, thought that would be a really good way for our team to share tips and tricks, and for people to be able to answer questions and then refer back to that information later on.” By integrating Teams with Slack, Bossert could tap into the way developers already asked each other questions, and give them a little nudge to start building Primer’s knowledge base. The team immediately saw organic growth of this repository without needing to overthink or assign documentation.

Primer has product teams that work on different versions of the service specialized for clients in verticals like finance, retail, and national security. Bossert describes how developers often need to reach across product lines. “Sometimes you might need to pull information from our news database for an experiment. It’s not something you’re familiar with or do everyday. Our teams can be siloed, but if someone needs a bit of code, that’s where Stack Overflow for Teams really shines for us.”

Onboarding - time to coding - development velocity

When developers use a similar tech stack or are working, as in Primer’s case, on different versions of the same product, there is often a shared set of tools. But there are still small variations between teams that create gaps in the knowledge and turn into pain points that reduce productivity. “There’s a legacy that has built up over time, a series of decisions on how exactly the tools are used.” says Stanley Fich, a front end developer. “It’s essential that all the decisions we have made in the past - the little configurations, the answers to how and why we structured our processes - can finally be found in one place. You spend less time looking for these, and more time actually coding.”

"When you get to the point where you can succinctly describe your problem, somebody's been there with you, whether you know it or not,” says Fich with a laugh. He estimates that Teams has helped Primer employees save hundreds of hours by avoiding common questions from new hires. "Being able to build up that knowledge base over time has severely decreased the amount of time it takes to ramp somebody up."

Identifying trends and experts within the organization

The benefits are not just about ease of work and an always up-to-date repository. There are noticeable changes in the company’s culture as well. Junior employees have learned that all opinions are welcome and that they may have knowledge which is valuable to the larger organization. "Using Stack Overflow for Teams has made me more confident,” says Du. “Now, if I see a question that I know the answer to, I don’t hesitate to offer thoughts.”

When someone has a question, instead of just dropping it into chat and hoping the right person sees it, they can search by tags to see who has contributed most on specific topics. “Say someone has answered 27 questions for a specific tag, you know they are probably the right person to talk to,” says Du.

The activity in Stack Overflow for Teams also offers a roadmap for how to structure DevOps going forward. For example, Bossert believes it has allowed him to spot trends that he might otherwise have missed. "Tags are a great indicator of which tools are used in your organization at any given time. If I see a new language or framework is gaining traction, then I know that might be something we want to invest in." In this way, developers help to lead from the bottom up, orienting the company around the tools that come naturally to them. "We can see if more people are using a new technology at Primer,” says Bossert, “And we can address those trends in policies or conversations about how our tech stack should evolve.”