Rob Witoff、Coinbaseプラットフォーム責任者 – インタビューシリーズ

Rob Witoff, Coinbase Head of Platform, brings one of the more unusual backgrounds in the crypto industry, spanning space systems engineering at NASA Jet Propulsion Laboratory, security engineering at Google, and multiple leadership roles at Coinbase and Polychain Capital. Before returning to Coinbase in late 2024 to help lead its platform strategy and expand onchain adoption, Witoff co‑founded institutional staking and cold storage company Unit 410, where he focused on infrastructure for institutional crypto custody and staking. Earlier in his career, he worked on navigation software for NASA missions including Dawn and Rosetta, and led systems engineering work tied to SpaceX’s first CRS mission to the International Space Station. His experience across aerospace systems, cybersecurity, cloud infrastructure, and blockchain architecture reflects the growing convergence between mission‑critical engineering disciplines and large‑scale decentralized financial infrastructure.

Coinbase is one of the world’s largest cryptocurrency platforms and the largest publicly traded crypto exchange in the United States, trading on the Nasdaq under the ticker COIN. The company operates across more than 100 countries globally and supports trading, custody, staking, and infrastructure services for hundreds of digital assets and cryptocurrencies. Coinbase previously disclosed more than 100 million verified users globally, though it has since shifted toward reporting monthly transacting users as a more active metric. The platform has expanded well beyond basic crypto trading into areas such as institutional custody, tokenization infrastructure, derivatives, payments, and onchain developer tools, positioning itself as a broader financial technology and blockchain infrastructure company.

You’ve had a unique journey from working on mission critical systems at NASA Jet Propulsion Laboratory to leading infrastructure, security, and now platform at Coinbase. How has that experience shaped your approach to building AI driven systems in crypto, where reliability, security, and real time performance are all critical?

When I was at JPL, I helped put a class 4 laser on the International Space Station. Our concerns weren’t only surviving the launch, but not blinding an astronaut in the process or causing an international incident.

To do that, we had to go through extensive safety and mission readiness processes with intense reviews and release gates. This isn’t too dissimilar to how we think about quality and security alongside a high‑performing system at Coinbase, we just have the opportunity to automate more of it here.

A couple of the lessons that I’ve taken from NASA to Coinbase are:

Stay focused on credible threats and have a robust way to think through them. We don’t want to boil the ocean. We want to have a framework to look through what we’re concerned about and burn through that list. Do we have a way to mitigate that risk? Do we understand the likelihood and impact of something going wrong? When we ship, we want to make sure we are protecting our users and our company.

Build in redundancy. In aerospace, we layer overlapping systems so one failure doesn’t bring everything down. Defense in depth.

Traditionally you’ll do that with multiple overlapping systems or multiple people. Now we’re able to do that with multiple models or systems with uncorrelated risks. We have different models checking each other’s work and multiple layers of protections, so even if one thing goes wrong, our users and our company are still safe.

Testing relentlessly. In aerospace, the integration and test phase, and the verification and validation phase where we extensively go through a lifelike scenario of all the things that we’re going to see on our launch, from vibration and thermal and vacuum testing.

We do the same thing at Coinbase. Whenever we’re shipping something to production, we invest heavily in the guardrails needed to ensure quality, which happens at several layers from unit, to functional to integration and finally staged production rollouts & observability..

We’re now doing that with prompts and agents. We’re able to codify a more diverse view on things that can go wrong from legal to compliance to security and make sure we’re reviewing and testing the changes we’re making before they hit production and giving our users the best possible experience. This isn’t always perfect, so we learn from edge cases and failures to build additional protections into our system so we’re always getting better.

The power of mission. NASA attracts incredible people in large part because of their mission. People are captivated by space and exploration. While our mission is different at Coinbase, we are similarly committed to it. People come here because we think the current system is broken, and we can do better. We believe we’re doing something meaningful for economic freedom and making finance accessible and more equitable for the world. 

When you have mission‑focused people that you get to work with, work is more fun, you can go further, you’re thinking bigger, and you can be really proud of what you’re doing with your colleagues.

Coinbase is focused on onboarding the next billion users onchain. Where is AI already making a measurable impact in simplifying the crypto experience for users, particularly around onboarding, custody, or transactions?

AI is transforming everything we do. Automating repeatable workflows, accelerating product development and enhancing human judgement are some of the biggest areas. 

Traditionally, workflows in customer support, fraud and risk investigations are some of our most expensive areas to scale while maintaining a high quality bar. Agents are transforming all of these, where we can provide clear goals + tools + context that are providing better, more consistent support experiences and rooting out bad activity faster so we can give all our customer a better experience.

From a platform perspective, what are the biggest technical challenges in scaling crypto infrastructure globally, and where is AI helping to solve problems that traditional systems could not?

The rate at which our infrastructure needs to scale up and down when crypto spikes can 10x or 100x our user demand in minutes because of a new token launch or change in public sentiment. That requires incredible elasticity and really well‑designed infrastructure, which is always a challenge at the scale and the security bar that we operate.

AI helps us get ahead of these spikes by spotting early signs of a sustained spike in traffic. AI is very good at distilling a large amount of information, looking at past events, and helping us predict what’s coming next. We’ve used that to successfully scale farther, faster, and significantly more efficiently, which has saved us millions of dollars and kept our systems healthy even under extreme load.

Separately, we support 50‑plus blockchain networks, each with their own upgrades and protocol changes. Keeping up used to be deeply manual, but now it’s now heavily automated. The same capability shows up in incident response, where AI helps us sift through information quickly, find the root cause, and fix it. That’s significantly reduced our mean time to resolution.

We have extensive use of agents and workflow automation all across the company, and as we safely scale up the services that Agents can talk to, we’re automating more of our work so we can focus on higher leverage problems.

There is a lot of hype around AI adoption, but in your experience, what separates companies that are truly embedding AI into production systems from those that are simply experimenting at the edges?

First is culture. You need to lead from the top, encourage people to delegate their work to agents and be hands‑on enough to know what you’re talking about. Such a radical shift in the way we work doesn’t happen by accident and takes dedicated time. You can’t win if leadership is delegating adoption to someone else.

Second, is investing in tools. The tools we used yesterday are not the tools we need today, and these tools change so quickly. We need to be really flexible in how we identify, pilot, onboard, and distribute tools in hours, not weeks. 

Then get out of your best people’s way. They’re already motivated, they’re already adopting the best tools, you don’t have to tell them how. Give them access to the latest models, the budget to use them, and the agency to work differently. Then let them run.

The space is only accelerating, and it takes persistence to not only keep up, but to lead.

Coinbase has long invested in developer productivity. How is AI changing the way internal teams build, test, and deploy services across such a complex platform?

Number one is making sure the platform isn’t complex—focusing on self‑service, composability, and building something that empowers engineers, doesn’t hold them back. That comes from simplicity and ease of understanding, which is a real investment.

How is it changing the way we build? Well, historically we worked with large teams because you needed a lot of people to cover a lot of ground. But large teams come with heavy ceremony & communication overhead to stay aligned. Smaller teams naturally have less overhead, and can now be augmented with agents to ship what used to take much larger teams.

From your vantage point, what does meaningful ROI from AI actually look like inside a crypto platform, and which use cases are delivering the strongest returns today?

We’re doing this all because we want to generate value, and no (good) company micromanages the value of every little thing they do. Instead, we hire people with good judgement, and give them the autonomy to act like owners and invest wisely with AI.

If you don’t have the right focus or strategy,, all of the AI‑assisted velocity in the world is not going to help you win. So we spend more time than ever ensuring we have these fundamentals right, so we can move fast in accomplishing the right goals.

Looking ahead, what role do you see AI playing in bridging the gap between traditional finance and crypto, and what still needs to be built at the platform layer to make that transition seamless?

This isn’t TradFi versus crypto, this is about what is the best way for people to control their finances and increase economic freedom. We think stablecoins are the right internet‑native way to pay, and distributed protocols are the right rails for them to move on.  Within Coinbase, we’ve seen a 100x growth in how we’re using agents across the company to do meaningful work to advance that vision, and it’s only growing from here.

Thank you for the great interview, readers who wish to learn more should visit Coinbase.