Founder spotlight: How PaxAI’s CEO tweaked work trials to snag top talent

Penny Chen doesn't believe in stylized coding interviews. Instead, the Pax AI co-founder and CEO runs two-to-three-day work trials. Candidates work in the actual codebase, build features that ship to production, and experience what it's really like to work at the company.

It's the same scrappy, practical approach that Penny brought to building Pax AI itself, which automates duty drawbacks, which is a US government program that refunds important tariffs. In less than a year since raising their seed round, the seven-person team crossed $1 million in booking revenue. 

But now they’ve faced a new and unusual challenge: they're turning away customers every week. And they’re not doing so because of product limitations, they’re doing so because they're at capacity. 

Penny's biggest regret is not hiring more aggressively sooner.

How Pax AI is unlocking tariff refunds for companies of all sizes

Penny met her co-founder Chris on Y Combinator's co-founder matching platform. They bonded over a shared passion for solving practical supply chain problems.

Before starting Pax AI, Penny was a research scientist at Flexport, where she learned about duty drawback firsthand. The program lets companies reclaim import duties when they re-export goods or use imported components in manufacturing. It's essentially free money, but only if you can navigate the complicated claims process.

The problem was simple. Traditional drawback providers take months to process claims, making it uneconomical to serve small and mid-sized businesses. Only billion-dollar multinationals could afford the service. 

Pax AI is building what Penny calls "TurboTax for duty drawback." The platform automates everything end-to-end, including data collection, extraction, calculation, and electronic claim submission. 

For larger companies with complex supply chains, work that used to take in-house teams months now takes days or weeks. 

For smaller companies with simpler cases, it's fully self-serve; you just have to upload your documents and you're done. But the impact is clear: SMBs can finally access refunds that were previously out of reach. 

Best of all, Pax AI's refund optimization algorithms ensure customers get the maximum refund possible by simulating all possible import-export combinations, which is something traditional brokers rarely do.

Play video3:49

How to run work trials that reveal the real fit

Most startups say they care about hiring. Penny proved it by rethinking the entire interview process.

Standard coding interviews involve one-hour sessions with stylized problems that candidates can prep for. She found that while these interviews are standard, they don't reveal how someone actually works — but work trials do. Here's how Pax AI runs them. 

Tip 1: Brainstorm the work together (treat it like their first day)

Penny doesn't hand candidates a pre-defined assignment. Instead, she treats the work trial like their literal first day on the job.

She gives context on the company, shares current strategy for GTM roles or the tech stack for engineers, and then they brainstorm how to design a productive day. 

That brainstorming process reveals how candidates think in uncertain, undefined settings, which is exactly what early-stage startup work looks like. You're not executing a polished roadmap. You're defining problems together, finding solutions together, and working through ambiguity in real time.

Tip 2: Give them access to your actual tools and codebase

For engineering candidates, Pax AI gives full access to the production codebase. Candidates can build real features that ship. They see the tech stack, the existing architecture, and the problems still left to solve.

This serves two purposes: 

1. It shows whether the candidate can actually do the work. 

2. It gets them excited about joining.

When engineers see you're using state-of-the-art AI tools and pushing the boundaries of what's possible in the industry, they want in. The work trial becomes a recruiting tool as much as an evaluation tool.

Tip 3: Shrink the timeline (one to two days is enough)

Pax AI originally ran week-long work trials. They quickly realized that was overkill.

"We can pretty quickly identify if it's a good fit within one to two days," Penny says. Shortening the timeline makes it easier for candidates to say yes, and it speeds up the hiring process without sacrificing signal.

The key is focus. You're not testing breadth, you're testing depth. Can this person collaborate? Do they ask good questions? Do they produce work that meets your bar?

Two days is enough to answer those questions.

What's next for Pax AI

Pax AI is hiring across engineering, operations, and go-to-market. But Penny's learned the hard way that finding great people takes time.

"If I were to go back, I would hire more aggressively," she says. The team is rejecting customers every week because they’re at capacity. Her advice: If you look at the data and the trajectory of growth, you can pre-hire to prepare for incoming demand.

It's a counterintuitive lesson for lean startups: Sometimes you need to hire before you feel the pain.

Though sometimes, the right solutions can help. Before founding Pax AI, Penny and Chris had never run HR. Rippling's PEO program gave them access to premium healthcare options at reasonable prices — critical for attracting talent in a competitive market. "We literally feel like we have an HR team supporting us," Penny says, pointing to the dedicated Slack support channel where their customer success rep responds within minutes.

For a seven-person team processing millions in refunds, that back-office support means Penny and Chris can focus on what matters: building product and closing customers.

And when you're turning away business because you're at capacity, every hour counts.