Max Liu

Profile

The Engineer Who Refused to Accept 'Database Scaling is Hard'

Before PingCAP existed, Max Liu was an infrastructure engineer at JD.com - China's largest B2C retailer, handling hundreds of millions of transactions. The database kept breaking. Not because of bad engineers, but because no database was built for the internet's actual scale. He watched teams burn months sharding MySQL, building brittle workarounds, praying nothing would go wrong during peak traffic. Liu decided to solve it at the source.

In 2015, he co-founded PingCAP with Edward Huang (CTO) and Cui Qiu, two engineers who shared the same frustration. The founding premise was stark: the world's most critical infrastructure - databases - had not kept pace with distributed systems. MySQL was 30 years old. NoSQL sacrificed SQL semantics. NewSQL attempts were either proprietary or not cloud-native. Liu and his co-founders sat down and wrote TiDB from scratch.

TiDB stands for "Ti Database" - a distributed SQL database that combines MySQL compatibility with horizontal scalability. It handles both transactions and analytics in a single system, a property the industry calls HTAP (Hybrid Transactional and Analytical Processing). The promise: your engineers never have to choose between consistency and scale again. Run your OLTP on the same cluster as your OLAP. Stop running double infrastructure.

"We are fundamentally an open source believer. TiDB, TiKV, and Chaos Mesh are collaborative achievements of developers and users, not solely PingCAP's products."

- Max Liu, PingCAP

This wasn't the first time Liu had bet on open source. Before founding PingCAP, he co-authored Codis - a Redis clustering solution that became one of the most widely deployed Redis management tools in China. The pattern had already set: identify a painful infrastructure problem, solve it in the open, let adoption prove the value. With TiDB, he scaled that playbook globally.

Building the Stack from the Ground Up

What makes TiDB structurally different from most database projects is that Liu's team didn't just build a query engine. They built the entire stack. TiKV - the underlying key-value storage layer - is a separate open-source project that graduated as a CNCF project alongside Kubernetes, Prometheus, and Envoy. Chaos Mesh, PingCAP's chaos engineering platform for testing distributed system resilience, also became a CNCF project. Three distinct open-source contributions to the cloud-native ecosystem, all originating from the same founding team in Sunnyvale.

The decision to open-source everything was philosophical, not tactical. Liu's operating theory holds that databases are infrastructure for the world's builders - and infrastructure becomes stronger when more people can break it, test it, and improve it. He specifically refuses to treat TiDB as "PingCAP's product." In his framing, TiDB belongs to the developers who use it and contribute to it. PingCAP just stewards it.

Key Insight

PingCAP's name is itself a distributed systems pun: "ping" from the network diagnostic command + "CAP" from the famous CAP theorem in distributed computing. It signals exactly who these founders are - engineers naming their company after a theorem most founders have never read.

Mission & Vision

From Sunnyvale to Global Infrastructure Standard

Liu's ambition is explicit and not vague: he wants PingCAP to become "the best and most respected infrastructure company in the world." Not a database company. An infrastructure company - the kind that quietly powers civilization's data layer while most people have never heard the name.

That framing shapes product decisions. When Liu talks about TiDB Cloud - the managed DBaaS version running on AWS, Azure, and GCP - he describes it less as a product and more as removing the last remaining friction between a developer and a world-class database. The engineering problem was solved in 2015. The operational problem is being solved now. What remains is making it feel effortless.

Under his leadership, TiDB has grown into a database platform used across fintech, gaming, e-commerce, logistics, and SaaS. The use case diversity is intentional. Liu believes the best signal that a database truly works is that radically different engineering teams keep choosing it for radically different problems. Pinterest stores social graph data on it. Plaid processes financial transactions. Gaming companies use it for leaderboards at massive concurrency. Each deployment teaches PingCAP something the product team couldn't have imagined in 2015.

"Our team's relentless drive for innovation and our laser focus on the needs of our customers and users" drives everything we build.

- Max Liu

In 2024, Liu keynoted HTAP Summit with a talk titled "What's Next for SQL @ Scale" - a signal that the category he helped define is now mature enough to have its own conference. The original insight, that transactional and analytical workloads should run on one system, went from unconventional to widely accepted in under a decade. The market moved toward Liu's thesis, not away from it.

The Operator Behind the Open-Source Idealist

Liu runs PingCAP with 600+ people across multiple continents. Revenue sits at $13.1M with ongoing investment in enterprise sales and cloud expansion. The business model follows the classic open-source playbook: the database is free, the cloud service and enterprise support are paid. What distinguishes PingCAP is how aggressively Liu has invested in the community as a product in itself. The TiDB community is not a marketing afterthought - it's a source of real product feedback, bug reports, and feature contributions that the commercial team couldn't replicate at speed.

His Twitter handle @ngaut1 is a quiet signal. "Ngaut" refers to an early Go library he authored. He still identifies as the engineer who wrote tools because they were missing, not as the executive who manages people who write tools. The identity has stayed consistent even as the company grew from three co-founders writing code in Sunnyvale to a globally recognized infrastructure platform.

In His Own Words

What Max Liu Believes

"A good database is not written, it's used."

On product development philosophy

"We are fundamentally an open source believer."

On PingCAP's core identity

"TiDB, TiKV, and Chaos Mesh are collaborative achievements of developers and users, not solely PingCAP's products."

On community ownership

"To become the best and most respected infrastructure company in the world."

On PingCAP's mission

"Our team's relentless drive for innovation and our laser focus on the needs of our customers and users."

On what drives PingCAP's success

"Enterprises were in a bind - choosing between scalability, performance, and management costs. TiDB removes that choice."

On TiDB's founding thesis

Context Clues

Details That Tell the Whole Story

🔗

His Twitter handle @ngaut1 references "ngaut" - a Go library Liu wrote early in his career. He still leads with engineer identity, not executive title.

⚙

PingCAP's name is a distributed systems pun: "ping" (network diagnostic) + "CAP" (the CAP theorem). Only engineers name their company after a formal computer science theorem.

📈

Before TiDB, Liu co-authored Codis - a Redis cluster tool that became one of China's most widely deployed infrastructure components. Open-source infrastructure is his native mode, not a pivot.

🌐

TiDB is deployed across every major cloud (AWS, Azure, GCP) and runs natively on Kubernetes. Liu's "run anywhere" philosophy was baked in from day one, before multi-cloud was mainstream.

💡

The team launched Chaos Mesh - a CNCF chaos engineering platform - from the same engineering culture that built TiDB. They dogfood failure testing at a level most enterprises only aspire to.

🔥

Dify, an AI application platform, reported an 80% infrastructure cost reduction after switching to TiDB Cloud - one of the most dramatic publicly reported database migration results.