Aman Khan

Excited to share I've joined Google Cloud AI, leading initiatives on our Agent Platform

After an incredible 4+ plus years at Arize AI building, scaling, and learning with an amazing team (so many people to tag but I have to pick just a few I know are online Jason Lopatecki Aparna Dhinakaran SallyAnn DeLucia Dat Daryl Ngo Matt Wilson), I can honestly say that I count myself lucky to have worked with this talented group of people. It's been an amazing ride and I can't wait to see what this team continues to ship.

Agent capabilities are scaling faster than anyone could have predicted - which is why managing Agents at scale (evals, governance, risk) has become vital for enterprises and their customers. Building agents that can do meaningful, impactful work requires ways to measure "Agent trust" (with tools designed for both humans as well as agents). This is a necessary and challenging problem space, with undefined solutions. I'm excited to be leaning in on this part of the stack with Michael Clark Michael Gerstenhaber Julia Wiesinger Brian Delahunty and many, many other incredible folks.

If you're at Google Cloud Next '26 next week, look me up!

2025 felt like a pretty wild year for me personally 🤯 I wanted to share a few learnings, anecdotes and resources from my journey with the hope that these reflections are helpful for others as well:

1. I transitioned from being an operator whose job was to find opportunity and align stakeholders, pitch an early vision, talk to customers and get stakeholders aligned (did I mention align stakeholders?) —> to embracing the identity of a builder, educator and strategist. A good portion of this year was spent thinking at a bigger picture of what our space (AI product evals) needed and how to educate people at scale. That required me to think about the journey I have been on for the last eight years of working in evals, and find ways to apply + teach those learnings. I've been in the eval space for a lot longer than a lot of the modern LLM systems account for, but it's because I've had this longstanding interest in black boxes and how they work that I've been able to take advantage of the capabilities that have emerged today. So if there's one takeaway here, it's to let your curiosity drive you. Almost no force I found is more powerful from a career perspective.

2. Just do the thing. Almost all of the outsized impact that I found myself being able to have this year came in the quiet time after meetings or before the day truly started, early in the morning or on the weekends. That quiet focus time was really where my best ideas and outputs came from. Unfortunately (and maybe to be expected), it took a while to see the initial gains of this work, with an interim period where I couldn't just drop all the commitments I already had with my team. This required close to ~six months of having two jobs at once until the fruit really started bearing. This feeling was familiar to when we went zero-to-one (and eventually one-to-one hundred) during our pivot into LLMs at Arize as well. The learning here is you can't just stop doing the things that you need to do. You just have to put in that extra work and carve out the time, because the extra 10-20% can lead to 2-3x outcomes. This is truly power law in a way.

3. We're still early. It feels like everyone on X this past week has just discovered Claude Code and the fact that you can build really powerful systems beyond just writing code. But the interfaces and how these systems work and what the right paradigms need to be is still early and not well-defined. In this messiness is opportunity. Even if you're just getting started today, I don't think that you're that much behind as you might think. And I think that's really exciting. More new perspectives allows the space to serve even more users and ultimately create products that are even more valuable. I want to see what you're building!

I'll link a few of the resources I mentioned above in the comments. Can't wait to share what's coming next this year and a few predictions in a future post 👀