Blog
Technical deep dives, product updates, and practical engineering insights for Data + AI teams
MatrixOrigin Recognized in the 2026 Gartner® Coolest Vendor Innovations in Data Management
MatrixOrigin has been recognized in the 2026 Gartner® Coolest Vendor Innovations in Data Management report as the only China-based vendor among the four companies named. MatrixOne Intelligence unifies databases, AI workflow orchestration, agent runtime observability, and persistent memory to provide a reliable, cost-effective foundation for enterprise AI applications.
MatrixOrigin Recognized in the 2026 Gartner® Coolest Vendor Innovations in Data Management
Good News | MatrixOrigin Named a 2026 Shenzhen "Potential Unicorn Enterprise"
Good News | MatrixOrigin Named a 2026 Shenzhen "Potential Unicorn Enterprise"
MatrixOrigin has been named a 2026 Shenzhen Potential Unicorn Enterprise, recognizing its technical innovation, growth potential, and market value in Data & AI.
MatrixOrigin Raises $10M Series A to Scale Enterprise AI Infrastructure
MatrixOrigin Raises $10M Series A to Scale Enterprise AI Infrastructure
MatrixOrigin has completed a Series A funding round of more than $10 million, backed by a HAND-led fund, AsiaCom, and Artesian Venture Partners. The funding will accelerate MatrixOne Intelligence development, global expansion, and enterprise AI deployment at scale.
MatrixOne Git4Data Deep Dive (Part 13) · Agents — Memory: From the Industry's Approaches to Governable Long-Term Memory
MatrixOne Git4Data Deep Dive (Part 13) · Agents — Memory: From the Industry's Approaches to Governable Long-Term Memory
Agent memory isn't a longer context window — it's the state layer that lets an agent keep working across tasks. This piece starts from what memory is, surveys the industry's approaches (prompt files, summarisation, vector retrieval, structured stores, platform built-ins), then shows how MatrixOne combines structured data, hybrid retrieval, and the branching, DIFF, snapshots and rollback of its Git4Data capability into long-term memory that is retrievable, auditable and recoverable.
MatrixOne Git4Data Deep Dive (Part 12) · Large Models — RLHF Preference Data: Disagreement, Adjudication, Reproducibility
MatrixOne Git4Data Deep Dive (Part 12) · Large Models — RLHF Preference Data: Disagreement, Adjudication, Reproducibility
Git4Data Part 12: preference data's unit is a pair, not a row, and it is computed from annotator votes rather than collected — the same data feeding both the RLHF and DPO routes. From 63,000 votes this derives preference pairs, audits degenerate pairs, no-consensus, preference cycles and length bias on a branch, materialises the conflict list before adjudicating on branches, and binds the dataset to its reward model. Verified on MatrixOne 4.1.0.
MatrixOne Git4Data Deep Dive (Part 11) · Large Models — SFT Data Curation: Auditable and Reproducible
MatrixOne Git4Data Deep Dive (Part 11) · Large Models — SFT Data Curation: Auditable and Reproducible
Git4Data Part 11: SFT data is orders of magnitude smaller than pretraining data, so every curation decision imprints on model behavior. Using one chat model's SFT pool, this runs a full curation pass on a zero-copy branch — exact dedup, near-dup, quality gate, safety, benchmark decontamination, multi-turn integrity — counting before each filter, with DATA BRANCH DIFF reporting the net change, then register-swap-snapshot to release. SQL verified on MatrixOne 4.1.0.
MatrixOne Git4Data Deep Dive (Part 10) · Deep Learning — Managing Training Data: lakeFS for the Files, MatrixOne for the Metadata
MatrixOne Git4Data Deep Dive (Part 10) · Deep Learning — Managing Training Data: lakeFS for the Files, MatrixOne for the Metadata
Git4Data Part 10: file-based (image, etc.) training data splits into two worlds — image/audio/video files go to lakeFS, the metadata (pointers, labels, hashes, splits) to MatrixOne. Training an image classifier as the example, do ingest, dedup, decontamination, integrity checks, relabeling, and curated release in SQL on the metadata, pinned by metadata snapshot × lakeFS commit; the end-to-end lakeFS+MatrixOne script is verified. SQL verified on MatrixOne 4.1.0.
MatrixOne Git4Data Deep Dive (Part 9) · AI Training in Practice — Dataset Release & Leakage: Don't Let Your Offline Metrics Fool You
MatrixOne Git4Data Deep Dive (Part 9) · AI Training in Practice — Dataset Release & Leakage: Don't Let Your Offline Metrics Fool You
Git4Data Part 9: the train/valid/test split is the step that decides whether your offline evaluation can be trusted. Using one risk model, detect and prevent five kinds of leakage (temporal, entity, duplicate, preprocessing, target) with SQL; then freeze the samples and the split manifest together with a database snapshot into a reproducible, auditable, reversible version, and compare the industry's other approaches. SQL verified on MatrixOne 4.1.0.
MatrixOne Git4Data Deep Dive (Part 8) · AI Training in Practice — From Data Arriving to Model Iteration: How the Git4Data Capability Runs Through the Whole ML Pipeline
MatrixOne Git4Data Deep Dive (Part 8) · AI Training in Practice — From Data Arriving to Model Iteration: How the Git4Data Capability Runs Through the Whole ML Pipeline
Git4Data Part 8, opening the AI-training arc: a coordinate system for the whole ML pipeline. From ingestion, quality gate, cleaning/labeling, feature engineering, and train/valid/test release, to evaluation, production monitoring, and feedback-driven retraining — what snapshot, branch, diff, merge, cherry-pick, restore, and PITR can and can't solve, threaded through one continuously-iterated risk model. Companion SQL verified on MatrixOne 4.1.0.
How We Rebuilt the Most Painful Part of Bidding with an AI Agent
How We Rebuilt the Most Painful Part of Bidding with an AI Agent
This article explains how MatrixOrigin rebuilt bid response workflows with an AI Agent, connecting tender Excel parsing, enterprise knowledge retrieval, evidence-backed technical responses, human review, and result backfilling into a verifiable, traceable, and continuously improvable process.
Agents Are Becoming the New "Smartphones" and "Smart Cars"
Agents Are Becoming the New "Smartphones" and "Smart Cars"
Agents are replaying the count curve of smartphones and smart cars — and will soon outnumber people. But the truly underrated asset is the Agent Trace each one produces: a petabyte-scale, high-intensity stream that today's data stack is forced to discard 86% of. Why Trace is the data flywheel for correction, memory, and training — and why enterprises need a data layer built to catch it.
MatrixOne Git4Data Deep Dive (Part 7) · Data Operations in Practice — Write-Audit-Publish: A Release Gate for Your ETL Pipeline
MatrixOne Git4Data Deep Dive (Part 7) · Data Operations in Practice — Write-Audit-Publish: A Release Gate for Your ETL Pipeline
Git4Data Part 7, closing the data-operations practice arc: Write-Audit-Publish (WAP). New data lands on a staging branch, passes a set of SQL audit assertions, then publishes with one atomic MERGE — a bad batch is stopped at the gate and production never sees it. With real scenarios, the full three-step walkthrough, and a detailed comparison against load-then-check / blue-green rename / staging+INSERT / transactions / DQ tools. All SQL verified on MatrixOne 4.0.0-rc3.
MatrixOne Git4Data Deep Dive (Part 6) · Data Operations in Practice — Collaborative Data Development: Merge Data the Way You Merge Code
MatrixOne Git4Data Deep Dive (Part 6) · Data Operations in Practice — Collaborative Data Development: Merge Data the Way You Merge Code
Part 6 of the MatrixOne Git4Data series, the Data-Ops theme: collaborative data development. When several people must edit the same table at once — parallel master-data curation, a reviewable data PR, developing a big change on a branch while mainline keeps serving — done with branch-per-engineer, row-level DIFF review, three-way MERGE, conflict policies (FAIL/SKIP/ACCEPT), and cherry-pick. Every statement verified on MatrixOne 4.0.0-rc3.