Live
Founder & CEO  •  NebulaGraph

Sherman
Ye

Building the graph database infrastructure for the AI era

Graph Database Open Source Enterprise AI Series A

Before founding NebulaGraph, Sherman Ye spent years building graph databases at Facebook and Ant Financial - the systems that mapped social connections and financial fraud patterns at planetary scale. In 2018, he decided to build his own. Today, NebulaGraph runs in the data stacks of Tencent, Meituan, and hundreds of other enterprises, processing trillions of edges at millisecond speed.

1T+
Edges Supported
200+
Enterprise Clients
$41M
Total Funding
120+
Employees
"While relational databases are capable of achieving many functions... they deteriorate in performance as the quantity of data grows."
- Sherman Ye, Founder & CEO

The graph database builder who got tired of building for others

There is a very specific kind of frustration that comes from working inside a system you know could be better. Sherman Ye spent the better part of a decade living with that frustration - first at Facebook, then at Ant Financial - engineering graph databases that handled some of the most complex, high-stakes data relationships on earth.

Ye joined Facebook in 2011, years before "graph database" was a category term Silicon Valley firms dropped into pitch decks. At Facebook, graph relationships were simply how the product worked - who knew whom, what they liked, how information traveled across the network. Ye worked on the infrastructure underneath it. Then Ant Financial, the Alibaba-backed fintech giant, recruited him to bring that expertise to financial data - fraud rings, transaction networks, credit relationships, millions of nodes updating in real time.

By 2018, the demand was obvious and the gap was real. Existing tools - relational databases pressed into graph duties, or proprietary graph systems with punishing licensing costs - couldn't keep pace. Ye left Ant Financial and founded vesoft Inc. The product was NebulaGraph: an open-source, distributed graph database built from the ground up for horizontal scale.

"You could still find relationships in data before, but relational databases become very slow as the data set grows."

Sherman Ye, TechCrunch, 2022

The bet on open source was deliberate and, at the time, risky. Enterprise database companies had historically guarded their core technology. Ye disagreed with that model. He open-sourced NebulaGraph on GitHub in May 2019 and let the community drive adoption. By December of that year, the repository had 1,614 stars and 269 forks. By 2020, it had 20 enterprise customers including Meituan and JD.com. By 2022, the community had grown from 60 users to 900+.

Ye's founding team reflected where he had recruited from: Alibaba, Facebook, Huawei, IBM. The collective resume wasn't assembled for optics - it was assembled because building a database that could handle trillions of edges without single points of failure requires people who have actually done that at scale.

Funding Journey
2020 — Pre-Series A $8M
2022 — Series A $20M+
Total Raised $41M
Key Investors
Jeneration Capital Matrix Partners China Redpoint China Source Code Capital
Tencent Meituan JD Digits Kuaishou Xiaohongshu JD.com 200+ enterprises

Use cases span fraud detection, recommendation systems, knowledge graphs, supply chain analysis, and social network modeling.

Twenty-five years of building at the frontier

~1999
Senior Developer at Atex Media Solutions - early foundation in enterprise software systems
~2006
Team / Tech Leader at LogLogic Inc. - log management and network security data infrastructure
~2009
Co-founded Taiji Networks - first entrepreneurial venture in networking technology
~2010
Senior Software Engineer at Corechange Inc. / Open Text - business intelligence software systems
2011
Joined Facebook as Software Engineer - four years building graph database infrastructure at social-network scale
2015
Joined Ant Financial as Senior Staff Engineer - led graph database work for financial fraud detection and credit networks
2018
Founded vesoft Inc. - launched NebulaGraph to bring enterprise-grade distributed graph database to the market
2019
Open-sourced NebulaGraph on GitHub - Alpha in May, 1,614 stars by December, 269 forks
2020
Closed $8M pre-Series A led by Redpoint China Ventures - 20+ enterprise clients including Meituan and JD.com
2022
Series A "in the low tens of millions" led by Jeneration Capital - community grew from 60 to 900+ users. Expanded beyond social/fintech into manufacturing, electric vehicles, and aerospace
2022
NebulaGraph joins LDBC (Linked Data Benchmark Council) to advance international graph database standards
2023
Pioneered GraphRAG concept jointly with LlamaIndex - merging knowledge graphs with large language models for enterprise AI applications
2024
NebulaGraph RAG platform announced at VLDB 2024. NebulaGraph Enterprise v5.0 delivers full ISO-GQL (standard graph query language) support
2026
Leading NebulaGraph with 200+ enterprise clients, 120+ employees, and expanding global presence from Cupertino, California

NebulaGraph: where trillions of edges go to work

NebulaGraph is an open-source, distributed graph database designed for one specific job: making very large, very complex networks of relationships fast and queryable. Where traditional relational databases require expensive multi-join operations that slow to a crawl as data grows, NebulaGraph keeps millisecond latency even when the dataset contains billions of vertices and trillions of edges.

The architecture is intentionally distributed - no single points of failure, horizontal scalability, automatic disaster recovery. The enterprise version supports ISO-GQL (the international standard graph query language), OpenCypher compatibility, hybrid cloud and on-premises deployment, and role-based access control for enterprise security.

The practical applications read like a catalog of where modern business complexity lives: fraud detection in financial networks, real-time recommendation engines, knowledge graphs for enterprise AI, social network analysis, supply chain mapping, anti-money laundering, credit scoring.

"With the ongoing digital transformation, more and more enterprises around the world have large linked data that consists of hundreds of billions of vertices and edges."

Sherman Ye, NebulaGraph

The GraphRAG initiative - launched in 2023 jointly with LlamaIndex - positioned NebulaGraph at the center of a new kind of AI application: one where a knowledge graph serves as the reasoning layer for a large language model. Instead of retrieving flat text chunks, GraphRAG retrieves connected entities and relationships. The result is more accurate, more contextual, and more explainable AI outputs. The whole thing can be set up in three lines of code.

NEBULA GRAPH Fraud KG Reco Social RAG AI GQL

NebulaGraph application surface across industries

Trillions of Edges
Millisecond latency at extreme scale - the benchmark set by Facebook and Ant
GraphRAG
Knowledge graph retrieval for LLMs - built in 3 lines of code
ISO-GQL
Full support for the international graph query language standard
Hybrid Deploy
Cloud, on-premises, or hybrid - with automatic disaster recovery

What Sherman Ye says about graphs, data, and the future

"While relational databases are capable of achieving many functions carried out by graph databases, they deteriorate in performance as the quantity of data grows."
TechCrunch, 2020
"You could still find relationships in data before, but relational databases become very slow as the data set grows."
TechCrunch, 2022
"Thanks to the in-depth understanding of industrial scenarios, the transformative value of our products, and the surging demand in graph technology, NebulaGraph is well-positioned to capture future growth opportunities."
Series A Announcement, September 2022
"With the ongoing digital transformation, more and more enterprises around the world have large linked data that consists of hundreds of billions of vertices and edges."
LDBC Membership Announcement

What the numbers show

Open Source from Day One
NebulaGraph was open-sourced in May 2019 - before "open-source database" was a standard enterprise go-to-market strategy. The GitHub community grew organically to 900+ enterprise users within three years.
📈
Community in Three Years
60 users to 900+ in two years. From 20 enterprise clients in 2020 to 200+ by 2026. The expansion moved from social media and fintech into manufacturing, electric vehicles, and aerospace.
🌞
GraphRAG Pioneer
NebulaGraph jointly introduced GraphRAG with LlamaIndex in August 2023 - an industry-first concept merging knowledge graphs with LLM retrieval. Now a platform at VLDB 2024 with sub-10-minute setup.
🏆
Standards Body Member
Joined LDBC (Linked Data Benchmark Council) to help shape international graph database benchmarking standards - a signal that NebulaGraph is playing in the long game, not just the product cycle.
💼
Marquee Client Roster
Tencent, Meituan, JD Digits, Kuaishou, and Xiaohongshu - these are not pilot customers. They run NebulaGraph at production scale for mission-critical applications in fraud, recommendations, and AI.
📄
ISO-GQL Support
NebulaGraph Enterprise v5.0 delivered full support for ISO-GQL, the international standard graph query language. Ye bet early on standards compliance as a competitive moat - especially for regulated industries.

Sherman Ye on camera

Interview Sherman Ye by The Next Database Platform
The Next Database Platform • September 2020
InfoQ Interview with vesoft Inc CEO Sherman Ye (叶小萌)
InfoQ • March 2021

The specifics that make it real

NebulaGraph's founding team was pulled from Alibaba, Facebook, Huawei, and IBM - not because it looked good in press releases, but because building a trillion-edge database requires people who have actually scaled data systems that size.

Ye's Chinese name is 叶小萌. VESoft's Chinese name, 欧若数网, doesn't obviously translate to "NebulaGraph" - the company operates under both identities across its China and global markets.

When NebulaGraph was open-sourced in May 2019, it had 1,614 GitHub stars and 269 forks by year-end. In the graph database space, that kind of organic early traction was a genuine signal - not a PR metric.

NebulaGraph's GraphRAG approach treats the knowledge graph as a "large-scale vocabulary" - entities and relationships as words. The whole system can be spun up with three lines of code, which is either impressive engineering or a very good abstraction, probably both.

Ye had planned to expand NebulaGraph into the US market in 2023. COVID-19 stalled the rollout. The company remains headquartered in Cupertino, California, with the bulk of its user base still concentrated in China - a deliberate bifurcation between engineering roots and market ambitions.

The company's competitors in the West are TigerGraph and Neo4j - both well-funded, both with significant head starts. Ye's differentiation from day one: open source, horizontal scale, and pricing that doesn't require a CFO's approval for a proof of concept.

What's happening now

July 2024
NebulaGraph Enterprise v5.0 - Full ISO-GQL
RC released with full-fledged GQL (ISO-standard graph query language) support - a major milestone for enterprise standards compliance.
August 2024
NebulaGraph RAG at VLDB 2024
NebulaGraph RAG platform announced at the Very Large Database conference - enabling GraphRAG applications for enterprise AI with minimal development overhead.
August 2023
GraphRAG Concept Launched with LlamaIndex
Jointly introduced the GraphRAG framework - industry-first combination of knowledge graph retrieval with LLM-based generation for more accurate and explainable AI.
September 2022
Series A Closed & LDBC Membership
Tens of millions raised in Series A led by Jeneration Capital. Simultaneously joined LDBC to advance international graph database benchmarking standards.

Where graph databases go from here

Ye's bet on graph databases predates the current AI moment by five years. The timing is not accidental - it's the result of having watched, up close, what happens when data relationships get too complex for tabular thinking.

The pivot to GraphRAG is the most significant signal of where Ye is steering NebulaGraph. As large language models become standard enterprise infrastructure, the question of how they access knowledge becomes critical. Vector databases retrieve similar text. GraphRAG retrieves connected facts. The difference in output quality is not marginal.

NebulaGraph's expanding footprint into manufacturing, electric vehicles, and aerospace - sectors where supply chain relationships and sensor networks create extremely dense graphs - suggests the market is widening faster than the product roadmap might have originally anticipated.

Ye remains Founder and CEO. The company has 120+ employees, is headquartered in Cupertino with operations in China, and has yet to make the full US market push that was delayed by the pandemic. That chapter looks increasingly inevitable.

"NebulaGraph is well-positioned to capture future growth opportunities."

Sherman Ye, September 2022