Skip to main content

LakeSoul: Cloud-Native Lakehouse Framework

Certified by the national information technology innovation program, LakeSoul serves as the digital foundation for enterprises to build a unified real-time data center and data asset base. It is also the data intelligence foundation for the AI 2.0 large-model era, supporting massive multimodal data management, integrated large-model training, and RAG applications

Star

Leading Technical Concept and Architecture

Traditional data architectures struggle with slow response times, high cost, and the inability to unify real-time and batch data at scale. LakeSoul provides a modern lakehouse storage layer that solves these problems, offering high-concurrency, high-throughput read/write capabilities and full warehouse management on the cloud, exposed uniformly to multiple compute engines.

Efficient, Scalable Catalog Metadata Service

Stores catalog metadata in PostgreSQL, improving metadata scalability and transaction concurrency.

Concurrent Writes and ACID Transactions

Provides concurrency control with high write concurrency, automatically detects and resolves conflicts, and ensures data consistency.

Incremental Writes and Upserts

Efficient merge-on-read and upsert capabilities improve data ingestion flexibility and performance.

Real-Time Lakehouse

Supports streaming and batch writes and snapshot reads; Flink CDC enables real-time multi-source ingestion with streaming incremental reads and computation, delivering a full end-to-end real-time data warehouse.

Multimodal Fusion Retrieval

Unified retrieval across vector search, table queries, and original files, with flexible filtering conditions and computation.

Open Ecosystem

Supports various compute engines such as Spark, Flink, Presto, PyTorch, and Ray, fully enabling data intelligence workloads such as real-time data warehousing, BI analytics, and AI model training.

Unified Stream-Batch Table Storage

Rich Application Scenarios That Unlock Business Value

Real-Time Data Ingestion into the Lakehouse

Flink CDC enables real-time ingestion directly from data sources — no T+1 batch imports or Kafka pipelines required

Real-Time Analytics on Operational Databases

With a few configurations for your online data sources, whole-database synchronization and real-time ingestion start automatically. New tables are detected and schema changes are synchronized without manual operations. Operational data is ingested into the lakehouse in real time and flows seamlessly into BI reports and dashboards, so key business metrics are always up to date to support decision-making.

Real-Time Report Analytics

Powered by unified stream-batch updates, data extraction and transformation are done in SQL, simplifying ETL and analytics.

RAG-Powered Intelligent Expert System

LakeSoul provides a native Python interface with support for mainstream AI frameworks such as PyTorch for direct data access, training, and inference. It also supports large-model training and RAG applications

GIFGIF
GIF

AI Expert Consultation

Join the Community and Share Data Intelligence