Ex-Alibaba team builds Asia’s answer to Databricks

Xiaohongshu processes an estimated 6 trillion browsing interactions per day from users scrolling through beauty tips, travel recommendations, and lifestyle content.

In the past, the company mostly relied on overnight batch processing, which means collecting a day’s worth of data and processing it in one go to build actionable user profiles. A customer could have 200 different attributes, from age and location to shopping preferences.

By the time Xiaohongshu’s data science team could use the data to deploy AI algorithms to personalize content and ads, the insights were already 24 hours old, says Ethan Yu, co-founder and CEO of Singdata.

According to Yu, Singdata’s cloud data platform lets Xiaohongshu analyze all 200 user profile metrics in near real time while lowering infrastructure costs by around a third. Clients “can actually profile the customers even before they leave the app,” he says.

Image credit: Arsal Ysfin

For Yu, a former Oracle and Alibaba Cloud executive, Xiaohongshu’s use case validates Singdata’s bet that current data platforms must process data almost instantly so AI systems can act on fresh insights without raising costs.

Founded in 2021 in Singapore, Singdata has raised funding from investors including Granite Asia, though Yu declined to provide further details.

The company has largely operated out of the spotlight despite a client list that includes NinjaVan, Ant Financial, and NeoCRM. Yu has likened Singdata to the “Databricks of Asia,” referring to the US data and AI company valued at about US$134 billion.

The CEO believes Singdata is well suited for customers running workloads across both US and China-based cloud platforms, allowing it to ride growing demand from Asian enterprises.

Built from scratch for Asia

A data platform is infrastructure software that sits on cloud servers, allowing companies to store, process, and analyze their own operational data. Think user browsing behavior, transaction records, logistics information, or customer interactions.

Historically, these platforms mainly turned raw data into structured formats for human analysts to query. But the shift to machine learning algorithms means raw data has become valuable inputs themselves, which eventually paved the way for a data lakehouse architecture.

This new model, which Databricks popularized in late 2020, combines the scalability and reliability of previous architectures (data warehouse and data lake). It lowers storage costs while supporting both analytics and AI workloads on the same foundation.

For Singdata, adopting a lakehouse architecture means customers can run traditional business intelligence and AI workflows on the same data layer. The platform bundles data integration, governance, and quality control into a single interface, similar in scope to Databricks’ Lakeflow platform.

What’s different, Yu notes, is Singdata’s cloud flexibility. In addition to three US cloud giants – AWS, Azure, and Google Cloud – the company supports four Chinese cloud providers: Alibaba Cloud, Huawei Cloud, Tencent Cloud, and BytePlus. Databricks, on the other hand, only works with Alibaba Cloud in China.

Different stake, different mindset

Can it become the Databricks of Asia?