Engineer Spotlight: Alan Xu on Transitioning from Microsoft, AI as a Partner, and Building Customer-Facing Features

Alan Xu joined Imprint as a Senior Software Engineer after three years at Microsoft working on infrastructure. In his first 100 days, he’s jumped into customer-facing feature development, contributed to a major partner conversion, and is now leading a project to drive card engagement through rewards.

What are some similarities that you’re finding between Microsoft and Imprint?

The two companies are very different in size, and I was also working on a very different type of product. At Microsoft, I spent three years mostly working remotely on an infrastructure product—the Software Load Balancer (SLB). It’s far removed from the end user; most people don’t even realize they’re hitting a load balancer.

At Imprint, I’m working directly on customer-facing features, which has been a big shift. That said, some things feel familiar. First, I’m surrounded by high-caliber engineers, which was one of the main reasons I joined Imprint. I get to learn and grow alongside them while building a great product that customers use every day.

Second, the problems are still complex—just in a different space. At Microsoft, scale was everything in distributed systems. Here, I’m building complex features for partners and customers. Third, there’s a constant push to improve: on-call processes, system reliability, and the product itself. If you’re excited about working on hard technical challenges and having ownership, Imprint is the place to be.

You joined Imprint at a complicated time, during one of our biggest launches. What have you learned so far?

I joined during the Crate and Barrel Holdings conversion, which is when a partner migrates from another platform to Imprint, and quickly learned that conversions are very much an all-hands-on-deck effort. It’s a full company push, with teams across engineering, product, and beyond focused on making them successful. They’re complicated and hard to execute, but they’re a big growth driver at Imprint. We did a full retro on our Crate and Barrel Holdings conversion to figure out what we can improve in 2026.

Joining during this period was especially interesting. I came in toward the tail end and immediately started building customer-facing features. I didn’t have much experience with React, but by partnering with Claude, learning from teammates, and reading documentation, I was able to help ship the product.

It was definitely challenging and a lot of work, but it’s critical to Imprint’s growth. I’m excited to continue improving our conversion process and what Imprint has planned for 2026.

Could you provide some insight on how you are using AI in your day-to-day work life? Do you find it overhyped? Is it actually helpful?

I find AI really helpful, but I treat it as a partner rather than a replacement. Having a strong foundation matters because it helps me sanity-check what AI generates and use it effectively. Instead of having it do everything for me, I use it to brainstorm my approach, build features incrementally, and double-check my work. That usually gets you much closer to what you actually want.

Most of my day-to-day work happens in Claude Code with MCPs. I use it to build features, triage issues when I’m on call, and speed up planning and debugging. Through MCPs, I can interact directly with Jira, Notion, Datadog, GitHub, and Figma, which makes it much easier to understand context and move quickly without constantly switching tools.

For code quality, I use Codex and Macroscope to review my pull requests. When I open a PR, they give me a first-pass review. I make the suggested changes and then send it to a teammate for final review, which helps reduce back-and-forth and keeps reviews focused on higher-level feedback.

I also use Notion AI to search through our documentation and catch up on context quickly, especially when I’m working in unfamiliar parts of the system.

Do I think AI is overhyped? A little, yeah. It’s really helpful and getting better all the time, but you can definitely lean on it too much. If someone doesn’t have much experience building applications and relies entirely on AI, they can run into problems pretty quickly. Having a strong foundation still makes a big difference.

What’s an interesting project you’ve worked on at Imprint so far, and what did you learn from it?

The first thing I worked on was getting the Crate and Barrel conversion across the finish line. I was mainly focused on the front-end, building our deferred interest UI, and spent my first month getting that shipped. During that time, we also did some application fixes and tackled tech debt, which was something we wanted to improve.

After that, I started leading a project called Streaks. Streaks is similar to Duolingo streaks, where users are encouraged to do something consistently over time. In Duolingo, that’s completing a lesson each day. At Imprint, we’re applying that idea to card usage by offering rewards based on customer spend. Right now, customers have a monthly spend goal. If they hit it, they receive a reward.

They can carry that streak into the next month and stack rewards as they keep using the card. It’s a great way to reward cardholders, give them great benefits, and encourage continued usage. For our team, the goal is to push our partners’ cards top of wallet, and Streaks plays a big role in that.