As organizations embrace modern DevOps practices, automation, infrastructure-as-code, and efficient deployment are key to boosting agility and performance. To explore these trends, we spoke with Anna Kennedy, an infrastructure expert! Anna is also one of our esteemed speakers at the Data Innovation Summit 2025!
In this interview, Anna shares her journey from physics to DevOps. She offers insights into scaling systems, common pitfalls, and best practices for teams at various stages of maturity. She also provides advice for organizations just beginning their DevOps journey. Learn more about how these practices can change your approach to DevOps!
Hyperight: Anna, you’ve had a career in infrastructure engineering. Can you share how your career journey led you to work in infrastructure engineering and DevOps?
Anna Kennedy: I studied physics and then computational geophysics at university thinking I might choose an academic career. But, along the way, it became obvious to me that I was much better at the computational bit than the geophysics bit.
So, after graduating, I took the common graduate path of joining an IT consultancy. After a bit of a false start in pre-sales, I moved to the Ops team. I was thrilled to be let loose on making things actually work!
From there, it’s been infrastructure of one flavour or another all the way. What’s been super exciting is to have been involved in the infrastructure space as it’s undergone an entire DevOps revolution. We have moved from entirely manual processes and editing live production code on physical servers to the world of automatic deployments and infrastructure-as-code cloud computing.
I think my natural curiosity and background of scientific training probably influences how I think about complex systems. When I need to debug a new workflow, I like to draw diagrams and then run experiments to ensure that my mental model matches reality. Without a clear understanding of a system, you can’t troubleshoot in any meaningful way.
Hyperight: Your presentation at the Data Innovation Summit 2025 will focus on modern DevOps practices like automation and deployment. Can you share how these strategies have helped reMarkable achieve faster and more reliable results?
Anna Kennedy: At reMarkable, I work in the data platform team, where we look after the common tasks and infrastructure that pull data from a range of sources each day and make it available for our analysts. We deploy most of our components with GitOps. That is, everything we run is described in code in a repo, and when that code is updated, we automatically deploy the resources to the right environment. This means that the development-to-deployment process is super fast and reliable, with a clear and auditable history.
One of the nice things about this process is that everyone can update the code as appropriate. So, analysts can make the changes they need and not be reliant on the platform team to move forward with their work. It also lets us set up modules and templates to provide a more intuitive interface for our internal users. This empowers them to provide more value to our business users.
Hyperight: For organizations just starting with DevOps, what are the first steps to implement automation and infrastructure as code? Especially if they are at an early maturity level?
Anna Kennedy: I think the first step I’d always recommend is to store everything as code in a repository, so that you can know what state the system is in, collaborate on changes, and confidently move forwards or rollback. From there you can think about automation and deployment, but having it all in code is fundamental. Even if you’re forced to make changes in a UI, most modern applications allow configuration to be exported as files, and that can be a lifesaver sometimes.
When it comes to automation, start small and add things incrementally as you gain confidence in the system. Choose tried and tested solutions where you have the most chance of community support. And avoid anything very niche where nobody will be able to help you troubleshoot if and when things get tricky.
Hyperight: As DevOps evolves, what do you find to be the most challenging barrier for organizations when it comes to adopting these practices? Particularly in terms of culture or resistance to change?
Anna Kennedy: Sometimes the resistance is purely a question of time and resources. For example, if a team is overloaded, they don’t have time to work on automating things, even if that would save a lot of time in the long run. Persuading the business to prioritize modernization can be an uphill battle, but definitely one that needs fighting.
I think fear can also cause a lot of resistance to change in many forms. Perhaps someone is worried about job security or losing their current authority. There might be a very real anxiety that replacing the old manual system with a new automated one will cause outages and breakage. I also think that it can be very uncomfortable being a beginner again when you’re used to being an expert. And there’s a lot to be said for working with psychological safety there.
Hyperight: In your experience, what are the common pitfalls when scaling DevOps processes across teams and ensuring that they remain effective at different maturity levels?
Anna Kennedy: One struggle I’ve come across is when a system is implemented in a way that is quite opaque to most users – it’s undocumented, unintuitive, and difficult to troubleshoot. In the long run, these systems tend to slowly go out of use because everyone avoids them and invents their get-arounds. I prefer modular workflows that can be tested at every step and upgraded piecemeal. And I love a good flow diagram!
Selecting which tasks to automate is also a common sticking point. Spending a whole sprint automating a process that takes five minutes a quarter might not be the best use of time. Unless you’ve finished everything else. I try to look for pain points, scary tasks, and easy wins, plus anything that involves a human remembering to do something (we’re not that good at that, or at least I’m not).
Hyperight: In your talk, you will mention implementing new processes into organizations. What’s one best practice you’ve found crucial when upgrading DevOps processes? Especially for teams that are more mature in their DevOps journey?
Anna Kennedy: It’s not the most exciting task, but sitting down and writing a rollout plan can be super helpful, especially for important and complex systems. For example:
Which components do you need to deploy to which environments and in what order? How will you verify each step in the upgrade and how long should you wait before moving on to the next one? What conditions constitute a rollout failure, and will you try to fix in place or rollback? If you’re rolling back, what will that look like in practical terms – a restoration of a snapshot, or a deployment of a previous commit or image?
Upgrades should be boring and routine, but it’s nice to be prepared in case they’re not.
Hyperight: Looking to the future, how do you see emerging technologies – like AI and machine learning – transforming the way we approach DevOps and automation in the next 5 to 10 years?
Anna Kennedy: I’m pleased that we’re already seeing products on the market that use ML to interpret monitoring data and alert on issues that have historically been a bit tricky to pin down without getting caught up in false alarms and corner cases. I’d like to see that expand into more self-healing – I’m sure AI doesn’t mind working on issues at 2 AM, but I do!
I’m also wondering if we will end up using AI as a compiler. So, instead of writing code in a particular programming language, you could just describe what you wanted in your native tongue and have AI compile it to a programming language. Or maybe just straight to machine code?
I think my biggest worry with AI is that we could end up with un-troubleshooting systems. Systems that are too opaque or complex for a human to ever be able to debug. I’m perfectly happy to let the computers take over the tasks that they’re good at and I’m not, but they will never be able to understand a system for me – unless we go for some kind of cybernetic solution I suppose!
If you’re looking to dive into the future of DevOps and automation, don’t miss Anna’s presentation at the Data Innovation Summit 2025! She’ll explore how automation, infrastructure-as-code, and GitOps are reshaping DevOps, sharing insights from reMarkable’s journey to faster, more reliable deployment. Anna will walk you through some best practices for scaling systems, overcoming common pitfalls, and creating a culture of continuous improvement.
Whether you’re just starting your DevOps journey or looking to optimize your existing processes, Anna’s session offers practical strategies that can help your team evolve. Join her at the summit to gain knowledge on building a more efficient and agile infrastructure!
Add comment