Today we’d like to introduce you to Paulina Xu.
Hi Paulina, please kick things off for us with an introduction to yourself and your story.
I started at MIT and, going in, I assumed I’d be there for four years like everyone else. But during my freshman year, I got much more interested in building things outside of class and working on startups. I found myself spending more and more time on that, and eventually it became pretty clear that it was what I wanted to focus on.
So I left MIT after my freshman year and went all-in on startups. Since then, I’ve spent a lot of time building around AI and figuring out where the biggest problems are as it moves from demos into real companies.
That’s ultimately what led to Agentic Fabriq. We kept seeing that once AI agents start actually doing things — accessing company data, using tools, acting on behalf of employees — a whole new set of problems shows up around permissions, security, and reliability. That became the problem we wanted to work on.
More recently, that also led us to build mnemiq, our open-source text-to-SQL engine, after seeing how hard it still is for agents to reliably work with real enterprise databases.
It definitely wasn’t a straight line from MIT to what I’m doing now, but I think the common thread has always been that I like building things and learning by actually putting them in front of people.
Alright, so let’s dig a little deeper into the story – has it been an easy path overall and if not, what were the challenges you’ve had to overcome?
The biggest challenge was probably the transition itself. I went pretty quickly from being a college student to being a founder, and that was a much bigger life change than I expected.
In school, there’s usually a pretty clear idea of what you’re supposed to be doing and whether you’re doing well. Starting a company is almost the opposite. There’s no syllabus, and a lot of the time you’re making decisions with incomplete information and learning from things not working.
I had to learn a lot very quickly — how to talk to customers, sell, hire, manage my time, deal with uncertainty, and just generally take responsibility for things I’d never had to think about before. There were definitely mistakes along the way, but that was also where most of the learning came from.
I think the biggest lesson has been getting comfortable with not knowing exactly what the right answer is, making the best decision you can with what you have, and adjusting quickly when you learn something new.
Alright, so let’s switch gears a bit and talk business. What should we know?
Agentic Fabriq helps companies safely connect AI agents to the tools, applications, and data they need to actually get work done.
The problem we kept seeing is that giving an AI agent access to a company’s systems is much more complicated than giving access to a normal employee or application. An agent may be acting on behalf of different users, touching sensitive data, and using multiple tools at once. Companies need a way to control what that agent can access, what actions it can take, and under whose permissions.
That’s what we’re building with Fabriq. We provide the identity, access, and governance layer between AI agents and the systems they interact with, so companies can adopt agents without giving them overly broad access or creating a new security problem.
What I’m most proud of is that we’re building around problems we’ve seen firsthand from companies trying to deploy agents in the real world. A lot of AI looks great in a demo, but production introduces questions around permissions, security, auditability, and control very quickly. We want Fabriq to make that part much easier.
Any big plans?
Right now, I’m most excited about continuing to build Agentic Fabriq and seeing how the way companies use AI changes over the next few years.
Agents are starting to move from experiments into real workflows, and I think that shift is going to create a lot of new problems that people are only beginning to run into. We want Fabriq to become a core part of how companies safely give those agents access to their systems, data, and users.
On a personal level, I’m also looking forward to continuing to grow into the role. I’ve already learned more in the last couple of years than I expected to, and I know there’s still a huge amount I don’t know yet. That’s probably the part I’m most excited about — continuing to build, meeting people who are much better than me at different things, and seeing where the company can go.
Contact Info:
- Website: https://www.agenticfabriq.com
- LinkedIn: https://www.linkedin.com/in/paulina-xu-5bba0821b/
- Twitter: https://x.com/paulinazhxu
