Andrew
Luo

The Engineer Who Automated the Job Everyone Hated

There is a Slack channel that launched a company. It lived inside Affinity, an investor CRM, and it was called "Data Importer Sucks." The people in it were customer success managers whose daily reality was cleaning messy spreadsheets before anything could be imported into the product. They were paid professionals drowning in inconsistent date formats. Andrew Luo read that channel and decided to fix it.

What seems like a mundane problem reveals its depth the moment you look closely. Every data field - a date, a dollar amount, a name - can arrive in dozens of variations. A date alone might be 10/10/23, October 10th, 2023-10-10, or a Unix timestamp. Writing code to handle even one field type correctly takes far longer than any engineer expects. What Affinity budgeted two months to build stretched into two full years of engineering resources. Andrew watched all of it happen, managed the team through it, and started wondering why every software company on the planet was building the same painful thing from scratch.

The Taiwan Decision

When Andrew left Affinity in early 2020 to start OneSchema with Christina Gilbert - a former Google product manager - the world promptly locked down. The US went into COVID restrictions. But Taiwan didn't. Coffee shops were open. Life was running. Andrew flew there and built the early version of OneSchema in a Taipei coffee shop while his home city was shut. That single logistical call - fly somewhere functional - is why the company exists. The alternative was months of stalled momentum at a stage where momentum is everything.

What they were building was an embeddable CSV importer: a component any SaaS company could drop into their product to handle all the intake complexity they'd otherwise build and maintain themselves. The product would validate data as it arrived, flag errors visually, let users fix problems inline, and deliver clean rows to the application. Simple premise. Genuinely hard to do well.

The Mockup That Sold Itself

Early pitches of OneSchema went badly. The concept didn't land. Andrew and Christina kept refining, kept showing different versions, kept watching prospects not quite get it. Then one mockup changed everything. They showed a spreadsheet with every data error highlighted in red. Instant recognition. An implementation manager interrupted mid-pitch - "When can I have this?" - and offered to pay for priority access before a working product existed. That moment is a lesson in the specific power of showing a concrete UI over explaining an abstract value proposition.

Getting into Y Combinator's Summer 2021 batch formalized what was already building. YC gave them the network, the validation, and an unforgettable marketing moment: they put their Product Hunt link on pizza boxes distributed at YC meetups. Everyone thought it was ridiculous. It landed them Toast as a customer - a restaurant technology company that genuinely understood the spreadsheet import problem from the inside.

Raising Without a Deck

In November 2022, OneSchema closed a $6.3 million seed round led by General Catalyst, with Sequoia Capital and Y Combinator participating. They raised it without a pitch deck. This detail is worth sitting with. A pitch deck is conventionally the central artifact of a fundraise - the leave-behind, the story container, the proof that you've organized your thinking. Andrew and Christina convinced two of venture capital's most selective firms to write checks anyway. The product, the traction, and the conversation were enough. Forbes noticed: both founders were named to the 30 Under 30 list in Enterprise Technology.

The company's customers as of 2024 included Toast, Vanta, Scale AI, and Ramp - a roster of fast-growing enterprise software companies that collectively show the breadth of the problem. Every category of B2B software eventually needs to ingest data from external spreadsheets. Every implementation team eventually builds or maintains an importer. OneSchema keeps showing up in those conversations.

The CSV Is Not Going Anywhere

In January 2025, Andrew joined the Data Engineering Podcast to make a case that sounds obvious once you hear it: CSVs will never die. The format is 50 years old, it requires no software to create, it is the lingua franca of every data handoff between organizations that don't share systems. The business world communicates in spreadsheets. The enterprise software world has to receive what the business world sends. OneSchema sits in that gap and has no plans to leave it.

The product has evolved since the original embeddable importer. OneSchema now handles recurring data feeds, PDF extraction, and AI-powered data normalization - moving from a point solution toward a broader platform for all the ways messy external data arrives at the door of an enterprise system. Andrew has a phrase for the aspiration: automate 90% of the workflow. The remaining 10% is the hard, context-specific judgment that humans still need to apply. Everything else should run itself.

What made Andrew Luo build this particular company is the same thing that makes the company hard to copy: he spent years inside the problem before he tried to solve it. He managed the engineers who suffered through it. He read the Slack channel where the people closest to customers said exactly what was wrong. The insight isn't that CSV imports are painful - everyone knows that. The insight is that the pain is so consistent, so predictable, and so widespread that it can be productized and sold. That is a different kind of engineering problem, and it turns out to be the most interesting kind.