.svg)
AI is changing the entire path into and through reliability engineering. Candidates are struggling to stand out, coding tests are losing their value, junior opportunities are narrowing, and the knowledge expected of an SRE keeps growing.
James Duffy is Principal SRE at Fabric. In this episode, he follows that career journey from the first hiring screen to the emerging discipline of AI platform engineering. He explains why he rebuilt his interview process around evaluating AI-generated Terraform and how that exercise surfaces production judgment instead of memorization.
The conversation explores how candidates can preserve a human voice in an AI-filtered job market, why junior hiring is essential to the future talent pipeline, and why continuous learning is the skill that connects today's interview to tomorrow's reliability work.
Key Topics Discussed
- (00:00) Why traditional engineering interviews no longer work
- (02:13) Testing production judgment with AI-generated Terraform
- (04:50) Standing out in an AI-filtered hiring process
- (09:32) Protecting the path from junior to senior engineer
- (15:59) Why the SRE learning path keeps getting harder
- (19:05) Continuous learning as a hiring signal
- (22:02) Learning into the next era of reliability work
Where to Find James Duffy
- LinkedIn: James Duffy
- Website: duffy.xyz
- Company: Fabric
Transcript
Why is the traditional engineering interview obsolete?
For most of my career, the interview process has been disconnected from the actual job. Candidates have been asked to whiteboard or complete tasks that do not reflect their day-to-day work. That tends to identify people who are good at interviewing, not necessarily people who are good at the job.
AI has broken even that imperfect signal. Almost anyone can now pass a simple programming test because AI can complete it faster than the best engineer. The old process was already flawed, but it worked just well enough. Now it fundamentally does not work.
How did you redesign the technical interview?
I first tried to make the interview resemble the job. It was open book, and candidates could research or use documentation while building something like a Terraform module. That was closer to real engineering work, but AI quickly made the exercise too easy. A task meant to take an hour could be finished in five minutes.
I changed the interview again. Instead of asking candidates to generate Terraform from memory, I give them pre-generated AI code and ask whether it is ready for production. It is not. Their job is to find and fix the integration gaps, security holes, and deprecated patterns.
At first, I let candidates generate the starting code with AI. As the models improved and failed less often, I switched to a fixed sample with known problems. The interview now tests whether someone can turn AI output into production-ready work.
How can candidates stand out when companies also use AI to filter applications?
The hardest part is getting the first interview. AI makes it easy to tailor a resume and cover letter, but relying on it too heavily can work against you. Many applications now share the same structure, formatting, and sentence patterns. If you squint, they look identical. A recruiter reviewing hundreds of them cannot remember one candidate from another.
The human touch matters more than ever. One useful approach is to give an AI the job description, your resume, and your cover letter, then ask it to act as a skeptical recruiter. Ask whether it would move you forward and what concerns it sees. Use that feedback to revise the application yourself. Do not paste the AI's rewrite back into the document. The feedback can come from AI, but the correction should sound like you.
What happens to the engineering learning path when companies stop hiring juniors?
Companies that stop investing in junior talent are not building a pipeline for future engineering teams. They will become top-heavy with senior people, and when those engineers move up or leave, nobody will be ready to fill their roles.
Every team should have at least one junior engineer so the team can continue to grow. Senior engineers are being asked to take on more with AI, which creates a burnout risk. Without junior people learning alongside them, there is nobody ready to replace that experience.
AI pricing is another reason to keep a balanced team. We may still be in the subsidized ride-share phase, when a service looks much cheaper than it will be later. AI companies are giving organizations large amounts of credit because they want adoption. As prices rise, companies may rediscover that junior engineers are effective at smaller tasks while AI is better reserved for difficult debugging and research. That balance lets junior engineers grow into the senior engineers the organization will eventually need.
Why is starting an SRE career harder today?
SRE and DevOps have always required broad knowledge. You cannot specialize in one technology when you are responsible for the reliability of everything running inside the company.
When I started, knowing four or five technologies deeply already felt difficult. Someone beginning today may be expected to understand 15 or 20 technologies inside and out, and that list keeps growing. I still think SRE is a great career, but I do not envy anyone starting from scratch now.
What should interviews reveal about a candidate's ability to learn?
You have to be a continuous learner. You cannot assume that what you learned in college or the tools you use today will still define your work in five years.
If I had been frozen in time at the start of my career, I would never have learned containers or Kubernetes because they did not exist in the form we use now. I had to learn those technologies quickly and become good at them. We do not know what the field will look like in five or ten years, and we are still in the early days of AI.
The strongest people I hire show a deeper drive to keep learning. They do not only learn something because a job requires it. They want to learn because the subject is interesting.
How will AI change reliability and platform engineering?
SRE and platform teams are beginning to absorb AI governance. The standards are still emerging, and providers expose different controls or use different approaches. A control that works well with one provider may not exist with three others, which makes consistent governance difficult.
Reliability and platform teams will have to take those emerging standards and create a consistent layer across providers. The work will include making AI capabilities safe and accessible while giving security leaders the controls they need.
What is AI platform engineering?
DevOps was a new term when I started my career, and much of that work is now described as platform engineering. I see AI platform engineering as an emerging subsection of that field. It is separate today because teams are building with new and poorly defined tools, but as standards mature, it will become part of platform engineering.
Traditional platform engineering grew around containers and Kubernetes. Platform teams made those complicated systems easier for product engineers who did not want to become experts in the underlying infrastructure.
AI platform engineering will do something similar around LLM gateways. As governance, controls, and logging become more complex, platform engineers will go deep into those systems and provide a safer, simpler foundation for product teams.
This is especially important in regulated healthcare. A public AI jailbreak may look funny online, but a similar failure inside a regulated product can do enormous damage. AI platform engineers will be responsible for the guardrails that reduce that risk.
What would you put on a billboard in San Francisco?
Your AI is not the risk. Your blind trust in it is.