Geneva International Invention Exhibition 2026
DeepEye

Steerable Self-Driving
Data Intelligence

A data intelligence agent system. One question, connect your data sources, and DeepEye automatically turns that into professional results.

DeepEye: One question, three professional outputs

Data is everywhere. Tools still treat it separately.

In real organizations, data lives in databases, Excel files, and internal systems — but most tools still treat these sources separately. People want AI that is not just powerful, but trustworthy, transparent, and auditable.

Gartner AI Spending Report
Market Scale · Gartner, Sep 2025

Worldwide AI Spending Will Total $1.5 Trillion in 2025

AI Application Software is forecast to grow from $83B to $270B by 2026 — a 3× expansion. The intelligent data analysis market is entering hypergrowth.

Read the full report ↗
EU AI Act Article 13
Regulatory Mandate · EU AI Act

High-Risk AI Must Be Transparent & Auditable by Law

Article 13 mandates AI systems be "sufficiently transparent." Black-box AI is no longer compliant — DeepEye's workflow engine meets this requirement.

Read Article 13 ↗
Harvard Business Review
Industry Pain Point · HBR, Sep 2025

AI Is Reinforcing Data Silos — Not Breaking Them

Current AI tools worsen data silos: each department deploys its own AI. DeepEye's unified orchestration breaks down these walls enterprise-wide.

Read the article ↗

Existing tools were not designed for real data work.

There are three big problems. Data is fragmented across formats. Complex tasks overwhelm single-agent systems. And many AI systems are still black boxes — users cannot clearly see or control the process.

The Heterogeneity Gap
Challenge 1

The Heterogeneity Gap

SQL, RAG, and spreadsheet tools exist separately. No single tool orchestrates all three — analysts must manually bridge the gaps.

The Context Explosion
Challenge 2

The Context Explosion

Single-agent systems overload one context window. As complexity grows, models hallucinate and forget constraints, producing unreliable results.

The Black Box Problem
Challenge 3

The Black Box Problem

AI agents are opaque: no plan inspection, no step validation. Sequential execution is slow and untrustworthy. You can't understand why results are wrong.

One question. Three professional outputs.

See how one question becomes a complete workflow — and turns into three professional outputs automatically.

DeepEye Full Demo

Analytical Report

Analytical Report

Executive summary, key findings, and trend analysis.
Not just formatted data — real written intelligence.

  • Executive summary and key findings auto-written and structured
  • Charts and insights connected to system conclusions
  • Draws on databases, Excel files, and JSON sources simultaneously
Auto-written Multi-source Export-ready

Interactive Dashboard

Interactive Dashboard

Filter, drill down, and explore the data directly.
No analyst needed — DeepEye generates it automatically.

  • Real-time filters update all charts instantly
  • Auto-generated trend lines, maps, and visualizations
  • Drill down from summary to individual data points instantly
Zero code Live data Real-time filters

Narrated Data Video

Data Video

Animated charts, narration, and subtitles — automatically.
One of our most distinctive features. Insights become easy to share.

  • Charts animate in real time with auto-generated narration
  • Synchronized subtitles frame-perfectly with audio
  • Professional broadcast-quality visual design
Auto-narrated Animated charts Broadcast quality

Simple for the user. Structured underneath.

The user types a question and connects data sources. DeepEye plans, validates, and executes — transparently, step by step.

DeepEye Workflow Engine
1
Type your question
Bind any data source with @ — databases, Excel files, or JSON data.
2
DeepEye plans the workflow
Decomposes into sub-tasks, builds a DAG, and schedules parallel execution automatically.
3
Validated before it runs
Every node compiled, validated, optimized. Errors caught before reaching your data.
4
Three outputs delivered
Dashboard, report, and video generated in parallel. Ready instantly.
5
You stay in control
Inspect, edit, and re-run any workflow. Drag a node to reorder — lines re-route and output updates instantly.

Three inventions. One coherent system.

Together, these three ideas form one coherent system — each addressing a fundamental limitation at the systems level.

The Heterogeneity Gap — data silos problem
Invention I
Unified Multimodal Orchestration

Unifies databases, Excel files, and JSON sources into one coherent pipeline. No silos, no manual integration.

SQL + NL2SQL Excel / JSON Multi-source Join
Context Explosion — single agent overload
Invention II
Hierarchical Reasoning with Context Isolation

Decomposes tasks into isolated sub-agents with focused context. Complexity no longer degrades accuracy.

Sub-agent Isolation Context Boundary Hallucination Resistant
Black Box — unreliable and slow AI systems
Invention III
Database-Inspired Workflow Engine

Treats workflows like database queries: compiled, validated, auditable. Parallel execution cuts hours to seconds.

Compile + Validate Parallel Execution Full Audit Trail

DeepEye improves over time.

It stores schema knowledge and successful workflows — reusing experience, retrieving the right context, and becoming smarter with continued use.

Document Memory
📄
Data Memory
Table schemas, field definitions, and business metadata indexed and auto-retrieved for relevant questions.
Schema Context
📌
Schema Context
Table names, field descriptions, and business definitions stored. DeepEye always knows your data.
Reusable Templates
Reusable Templates
Successful workflows saved as templates. One click re-runs instantly.

Not just a demo — built for real deployment.

We are not only thinking about intelligence. We are also thinking about security, control, and enterprise use.

🐳
One-Click Docker Deployment
Deploy the entire stack with one Docker command — on-premise or any cloud.
🔒
Sandboxed Code Execution
All generated code runs in isolated containers — no unauthorized access to the host system or network.
👤
User Authentication & Access Control
Role-based access control — analysts only see authorized data sources, every query attributed to a user.
Parallel Execution Engine
Independent nodes run simultaneously via topological scheduling — turning sequential hours into parallel seconds.
🌐
Unified Multi-Source Integration
Connect PostgreSQL, MySQL, Excel, JSON, and REST APIs in one session — no custom connectors needed.
📋
Fully Auditable Workflow History
Every query, decision, and output is logged with full provenance — every result traceable to its source.

Researched, protected, and validated.

DeepEye has received strong external recognition — through awards, patents, and academic publication.

AI Agent 2025 Best Open Source Project Award
Competition Award
Best Open Source Project
AI Agent 2025
Best Project Award
5 Patents Granted
Intellectual Property
5 Patents Granted
China National Intellectual Property Administration
Autonomous Data Analysis Method & System
DeepEye Demo Paper
Academic Publications
10+ Publications
Published in top-tier AI and database conferences,
including SIGMOD, VLDB, ICLR, ICML and NeurIPS.
DeepEye Logo DeepEye
HKUST(GZ) Logo
Geneva International Invention Exhibition 2026
Steerable Self-Driving
Data Intelligence
The Hong Kong University of Science and Technology (Guangzhou)
Scan to visit DeepEye
Scan to explore online
xypkent.github.io/deepeye