Ran Chen, CTO of Pure Global and VayoMed, on building AI businesses, taking a career break for his family, and why finishing projects faster made finding customers urgent.

Ran Chen is CTO of Pure Global and VayoMed. Pure Global uses AI to streamline medical device registration, while VayoMed helps life sciences companies with marketing and customer acquisition. Ran entered Tsinghua University's School of Software in 2009. Then he went to Carnegie Mellon University (CMU) for a master's in computational data science, focusing on machine learning. After graduating in 2014, he worked on applied machine learning at real estate platform Trulia, then joined free streaming service Tubi in 2017 as its first ML engineer and later became an engineering director. He subsequently spent about two years as a stay-at-home dad, studying financial planning and earning professional licenses during that break.

In 2025, Chen joined Pure Global, where his team used AI to rewrite internal systems and built a shared knowledge base. As the team began completing projects faster, finding more customers became its next challenge, leading it to incubate VayoMed, which helps life sciences companies with marketing and business growth. His side projects include Open Exam Preparation, a free study website inspired by his own exam preparation. Across both his side projects and his company work, Chen keeps asking the same question: whose problem does AI solve, and how does it help the business run better?

In our in-depth conversation, we discuss:

  1. Why Chen chose machine learning roles with measurable business impact, and what Trulia and Tubi taught him about putting algorithms into production
  2. How two years as a stay-at-home dad changed Chen's thinking about family responsibilities, wealth, risk, and life choices
  3. How a free exam-preparation site reached about $10,000 in monthly revenue, and why model and labor costs matter
  4. How Chen uses search data to find unmet demand and revenue feedback to decide which AI projects to pursue
  5. How Pure Global rewrote its systems and built shared knowledge, shifting its main constraint from delivery to customer acquisition
  6. How AI helped Pure Global lower its fees, and why service pricing and employee incentives also need to change

My biggest takeaways from this conversation:

  1. Machine learning only ships when the rest of the company is ready to change.

    Chen wanted to work on technology whose business impact could be measured: whether an algorithm increased revenue or made a new service possible. At Trulia, however, running experiments and deploying algorithms also meant getting time on other teams' schedules, which was difficult without support from senior leaders. At Tubi, where he built the data and machine learning systems from scratch, his management priorities included defining how to measure results and removing obstacles between teams. His pitch when looking for a job was direct: "I'm here to put AI to work for you." That requires engineers to take ownership — and companies to give them the conditions to succeed.

  2. Chen stepped away from work to care for his children and give his wife room to build her career.

    When Chen left Tubi, his two children were still young, and his wife was working at a Big Four accounting firm while studying for a part-time law degree. He had some savings and wanted time to consider his next step, so he took on more of the childcare to help her continue studying and advance at work. The payout from Tubi made a few years off possible, but not permanent retirement. His view of money explains the choice: "We don't make money just to make more money. We make it so we can use it when it matters." He also began studying financial planning during this time, to understand how to manage what he had earned.

  3. A tech salary, a Silicon Valley home, and employer stock can all rise and fall together.

    Learning about financial planning showed Chen how differently risk looked to people in Silicon Valley and to professional planners. A family's income came from tech, its stock came from an employer, and its home value depended on the local tech economy. One industry downturn could hit several major assets and income sources at once. Chen acknowledges that people who held company stock for years sometimes made fortunes, but he doesn't think that settles whether taking that risk makes sense for a particular person. He uses a lottery analogy: "If you win the lottery, do you spend it on another lottery ticket, or go and enjoy your life?" Once you have wealth, you still have to decide what risks you can tolerate—and what life that money is meant to support.

  4. His free exam site runs on ads, but it stays alive by keeping costs down.

    While studying for professional exams, Chen bought plenty of materials but often couldn't find explanations he liked. He ended up asking AI and hunting down resources himself. That experience became Open Exam Preparation, a free site that offers study materials across a growing range of exams. He told us it brings in about $10,000 a month, mostly from advertising, plus referral links to books and paid exam platforms. When asked why someone couldn't just copy it with AI, he pointed to starting early—and to costs. The trick is choosing the right models, using fewer tokens, and limiting human labor so the service stays cheap to run. The project is built to run without employees, but it still needs oversight: "I check on it every day. But it doesn't really need much attention."

  5. Look for unmet demand before deciding what to build.

    A food blogger who reviews struggling restaurants reminded Chen of a mistake he sees among tech founders: opening a bakery because you love bread, without figuring out where the customers are or what they want. For his online projects, Chen digs into search data to find queries people are making but not getting good answers to. Then he decides what content or service to offer. "What do those people want? That's what you offer." He draws a line between two goals. If you're raising money to explore something no one has built, you can accept a stretch with no revenue. If you want a profitable business, you start by calculating revenue and expenses—and understanding why customers would pick you over someone else.

  6. AI makes experiments cheaper. Real feedback still decides what survives.

    Chen used Suno to make music and put it on Spotify and Apple Music, but stopped investing time because it took too long to see results. At the time of the interview, it was still bringing in a little over $100 a month. He also uses AI to surface project ideas, builds them, watches whether they make money, and feeds those results back into the next search. Because building and running a website costs less now, he can test several ideas without much capital. He has come up with ten to twenty candidates, but expects most to fail. The exam site was different: he knew from experience how hard it was to find useful materials and answers, so he knew what to improve—and was willing to push through the period before it paid off. Building a website is easy now; understanding its users still takes time.

  7. Rebuilding with AI also means deleting features customers don't need.

    At Pure Global, the turning point for Chen was AI becoming part of the workflow and helping the team finish tasks. He used AI to rewrite internal systems, spent time talking to customers, and cut many existing features: "It turned out that what customers wanted was much simpler than what we'd been offering." In his view, some features had been built because teams wanted to pursue their own ideas or justify their existence, rather than because customers needed them. The rewrite retained a smaller set of useful functions and made the systems easier for both employees and AI to use. It was also a chance to reconsider which work was worth doing.

  8. When AI gives a bad answer, fix the knowledge base.

    At Pure Global, a lot of working knowledge used to live in employees' heads. New hires had to track down the right person to ask. Chen's team spent a long time building an internal knowledge base and encouraged people to ask AI first. He says, "If the AI gives a bad answer, we fix the knowledge base." Each poor answer gives the team a chance to improve the knowledge base, so the next employee with the same question gets a better answer. Chen says the knowledge base has helped new colleagues get up to speed faster. Organizing that knowledge — and getting AI to use it well — still takes ongoing effort.

  9. Faster delivery let Pure Global cut fees and focus on winning customers.

    Pure Global helps medical device companies register products with local regulators through nearly 20 offices worldwide. AI-assisted document writing helped the team clear a backlog of 50 to 60 projects. Chen says its delivery capacity now far exceeds its ability to acquire customers. As registration work became cheaper to deliver, the company published its prices and stopped charging a separate registration service fee, moving to a flat annual fee. With spare delivery capacity, the team turned its attention to customer acquisition and incubated VayoMed, which provides growth services to life sciences companies. VayoMed is meant to solve Pure Global's own customer acquisition problem—and help other companies with the same need.

  10. Chen wants employees to share in the extra income AI helps generate.

    If AI saves time and only shareholders benefit while employees worry about being replaced, Chen argues, workers have little reason to look for better tools. He says the company has already changed some incentives and hopes to see more partnership-style arrangements. When the extra income goes to the people doing the work, they have more reason to improve services, adopt new tools, and bring in more customers. That's the organizational change he wants: people who take on responsibility should share in the additional income their work generates.