>

Base Layer EP 06: Ari Shamash on Your Agent Shouldn't Have Your Password

[Listen now]

Q&A: Being an SRE at AuthZed

Two members of AuthZed's SRE team on what the work looks like, what they own, and why they chose AuthZed.

Jenessa Petersen
November 12, 2026·10 min read

SRE at AuthZed means running authorization infrastructure that sits on the critical path for every customer, while redesigning that infrastructure for the next order of magnitude of growth. We sat down with two members of the SRE team, Quinn Murphy and Lu, to talk about what the work actually looks like, what they own, and why they chose AuthZed.

Answers have been edited only to remove filler words and false starts.

What does a typical week look like?

Quinn: Lately I've been delving into one of our larger customers. They've got things in a bit of a bespoke environment for us right now, and so I've been spending a lot of time there untangling, reconciling our processes, what we are normally doing with what we have to do there. But then also, as I'm going along, trying to resolve a good amount of technical debt that's been caused by the difference there, and just trying to bring them up to our current standards. It's been a pretty interesting project.

Lu: I think one of the cool things that you're doing there is that you're also taking this opportunity to automate things in a way that's maybe not the end-all, be-all best thing that's aligned with all of our other environments, but it's just really pragmatic, and it makes our lives better in the day-to-day. How can I get 80% of the way there in just 20% of the effort?

Every week, we have a mixture of planned quarterly goals that we're making incremental progress towards, and then we're also handling a lot of interrupts. We're working directly with customer requests, like, oh, we want to enable some special features to optimize the performance of SpiceDB. And then also just debugging problems that arise at the scale at which we're operating. We have this really big mesh of a ton of environments, and just by rolling the dice so many times, you end up hitting weird situations.

And then in parallel, I'm having some AI tools improve our internal CLI and our TUIs to make our rollouts and our automations better. So we're always doing some things that we want to do, some things that have to get done, and then just some things that make our lives better.

Quinn: Yeah, I want to say kudos. I've been using one of those tools that you built as I'm provisioning today, and loving it. It's just much better cognitive ergonomics. I think we've been doing a good job of keeping the lights on and building a path forward to the future.

How much of the job is automation, and is that exciting?

Lu: For me, it's definitely really exciting. We're a startup, and we grew really fast, so we did some initial amount of automation early in the day, and then we went with that automation as far as we could. We've been struggling a little bit under the weight of that original design for the last year, because our initial design didn't really account for this future. So now we're working on improving our workflows every single day to have a system that is able to support the next order of magnitude greater. It will liberate us to work on stuff that is way more intentional and is able to deliver new value for the company for the next 2, 3, 4, 5 years.

Quinn: For me, a good model for SRE is looking for new problems, right? Problems are never going away, and friction is always there, but you can always be upgrading your problems. We originally had the problem of how do we get customers? And we solved that problem, and now we have a bunch of customers. Our new problem that we've upgraded to is how do we maintain them easier? And automation is often the path of upgrading your problems. You not only can defer the actual fingers-on-keyboard work, but you can defer the thinking-about-it work, which is often the big toil of it. When you automate something, you're taking what you know and locking it in there and putting it to work.

Lu: We have on the order of hundreds of environments right now, and one of the things that I'm really excited about having automated in the last couple of months is a bunch of static validation tooling. It's able to translate these really loose type annotations in a language that is dynamically typed, and use those to statically analyze tons of manifests, and automate all this CI checking that we would have to do manually. Now we can defer all of that to the computers, and we can focus on cooler things.

How much of the reliability playbook was written when you joined, and how much is left to write?

Quinn: I think a decent portion of the reliability playbook was done in certain places, but then, because we had to move so fast, there's that drift. When I first got in, my main emphasis was reconciling what I'm reading with what's there. Now we're moving forward, and for parts that we as a team hadn't gotten a chance to look at, we're getting a chance to build out that part of the playbook, which is exciting.

Lu: When I joined the team, the team already had a really strong foundation of reliability for the technology that we were already deploying. Since then, we've gone through multiple iterations of new technological evolutions, and all of these are going to require new reliability playbooks and new deep dives. Right now we're going through a really big internal effort to evolve our product for the next generation of SpiceDB and AuthZed. That's going to be a really big undertaking for the SRE team, to flesh out what this playbook looks like for the next 5 or 10 years.

Is SRE here just about reliability?

Lu: The SRE team isn't just responsible for reliability and observability, but is also responsible for maintaining and evolving the architecture for how we run all of our infrastructure. And we're just starting to set sail on a new direction, which is, cool, how can we do our internal architecture V2? To reduce costs, make things faster, have better end-to-end tests, better visibility into everything, rollbacks, maintenance, all of this. There are so many big internal initiatives that we're really excited to work on, and we just want more people to help us work on it.

What ownership do you have here that you might not get at a bigger company?

Quinn: Maybe a month and a half after I'd started, we had one of our newer features on a new platform, and we were trying to do a proof of concept. It was also in a rush for a customer, and I'm not sure if, in a more established place, I would have been tasked with that. There's that need, in a fast-paced place, to just pick up, take ownership, and drive. I had the team there to answer questions as I needed them, and then I just grabbed it and went.

Lu: Since you joined, we had a huge effort, which was, alright, everything we've built, let's deploy all of that to a new cloud. We run managed infrastructure for our customers across a number of different hyperscalers, and we added a new one recently. That's something that, in a big organization, just doesn't happen on a Tuesday. But we have a big enough customer that wants this, and it's like, cool, let's go figure out how to make this work. We'll be working really directly with product and with customers, and there's something really rewarding about that.

Authorization sits on the critical path for every customer. How does that change how you think about latency and availability?

Quinn: When your customers rely on you constantly for the majority of their work, you are not just a SaaS, you are infra. With a SaaS, the hidden feature that SREs deal with is typically uptime. If it's not up, then none of the other features matter. For us, obviously we want to keep the infra up for our customers, but performance and latency are actually going to be bigger deals. If we're slow, now all of their stuff is slow. We have to be almost performant-first, so that requires us to be a little bit more proactive.

Lu: I've also worked in smaller startups that weren't shipping such mission-critical things, so foundational to a customer's environment, and it meant that we could ship really quickly and be really YOLO. Here, we need to be really careful with not breaking things. We could easily find ourselves between a rock and a hard place: we can't ship things so quickly, but we also have a lot of customer demand, so we need to ship things quickly, and we are a small startup without a lot of resources.

What made you say yes to AuthZed, and what has kept you here?

Lu: The biggest thing that made me say yes, for sure, was the team. I've known a lot of the team for almost a decade, and I knew for sure that the people that I did know were really smart, really empathetic, and just really effective teammates. I had all the confidence that no matter who they've added to the team, they're also going to be similarly effective and nice and great to work with. That's also what's kept me here: this is just a really good team of smart and effective people that are a joy to work with every day.

Quinn: I didn't know a lot of the people here beforehand, but I will agree with the effective, empathetic stuff for sure. I like working places where I feel we're solving a real problem, and our general mission has a real impact on how things are done. And when I was talking with people in the interviewing process, everybody just seemed awesome to work with. Some of the best conversations I'd had in the interviewing process. What keeps me here is lots of opportunities for great ownership of things. We've got a pretty dense stack, so there's lots of stuff to dig into, just constant learning and moving. I really dig that.

If an SRE were on the fence about applying, what would you tell them?

Quinn: On the fence? Impossible! If you're looking for a place where there's not just high ownership, but high ability to drive direction and make change, and to work in a good culture, this is really your place. Get off the fence, get over here.

Lu: Yeah, I'd say that if you're looking for a place to chill, this is not the place. But if you're looking for a place to really be able to drive and make a cool impact with a lot of nice people, then definitely come here.

Join us

AuthZed's SRE team is building the next generation of infrastructure behind SpiceDB, and there's a lot of it left to design. If building it, not just running it, sounds like your kind of work, check out our open roles.

Related

See AuthZed in action

Build delightful, secure application experiences with AuthZed.