Insights from a PhD in Databases: What Time Taught Me About Building Infra

This month (April 2025) marks 4 years and 1 month since I started building RisingWave.

GitHub: https://github.com/risingwavelabs/risingwave

Hard to believe we not only survived — but actually thrived — growing faster than we expected every single month. Sustaining high growth is never easy for a data infra company, especially during the peak of AI hype. That’s something I’ve never taken for granted. Looking back, here are a few lessons time has taught me.

Make friends with time

Building data infra is brutal. No matter how hard you work, the first few years feel like shouting into the void. People don’t trust you — not because you’re wrong, but because you’re “too early.” But if time is on your side, and you believe in what you’re building, stay the course.

When we started RisingWave, our core principle was clear: democratize stream processing with Postgres compatibility and S3 as the primary storage. Four years later, I still believe this design will stand the test of time.

And above all, cherish every customer who trusted you before it was obvious.

Be water

Adapt.

When the AI boom hit, we took a hard look at everything: Is data infra still something the market needs? Should we pivot to AI — chatbots, agents, the works?

Our answer: we’re proud to be a data infra company. But that didn’t mean standing still.

We tore down our GTM motion and rebuilt it from scratch. No more Google ads. Instead, we embedded AI in our marketing and sales workflows. We bet on high-quality content and GitHub presence. If we had kept chasing growth the traditional way, our ROI would’ve collapsed.

Ride the wave

Every few months, I ask myself: what’s next — 3 months, 6 months, a year out?

Two years ago, the answer was simple: be the best stream processing engine. I think we’ve done that.

But the game changed. Iceberg has become the default table format. Everyone wants to build an Iceberg-based lakehouse — so we went all-in.

We’re not building a monolithic, all-in-one lakehouse. We bet on openness. Our job is to help users lower the bar for building their own lakehouse. We handle ingestion, transformation, and storage optimization — users bring whatever query engine they like.

Today, Iceberg is RisingWave’s #1 data destination — and our visibility is growing fast because of it.

Conclusion

Looking back, I feel lucky. But luck doesn’t show up unless you stick around, stay consistent, and stay open to change.

So if you’re building data infra — or stuck in an AI anxiety spiral in 2025 — don’t panic.

Stay long-term. Stay flexible. Ride the wave.