Data Intelligence Agents (DIA): Autonomous Coding Agents for Enterprise Data Integration and SQL Query Generation

Forum topic · 小凯 · 2026-06-19

Summary

Data Intelligence Agents (DIA) is a three-agent system—Data Interpreter, Schema Creator, and Query Generator—that streamlines production data integration, which is typically bottlenecked by repeated, lossy handoffs between data owners, engineers, and analysts. Built on autonomous coding agents (ACAs) as a first-class abstraction, DIA's agents generate, execute, validate, and repair concrete artifacts instead of merely emitting text, reuse experience via shared memory, and surface results for review by domain experts. The system is deployed in production for enterprise customers. The authors evaluate the Query Generator in fully autonomous mode across seven SQL benchmarks spanning four task categories and four dialects, where it matches or surpasses the best published results on all seven. This demonstrates that an execution-grounded architecture built on ACAs and shared memory can generalize across data intelligence workloads, with adaptation confined to natural-language instructions. Paper: arXiv 2506.14970.

Paper Overview

Research Area: Machine Learning (ML)
Authors: Anoushka Vyas, Aarushi Dhanuka, Sina Khoshfetrat Pakazad
Published: 2026-06-19
arXiv: 2506.14970

Problem

Production data integration is bottlenecked by repeated, lossy handoffs between data owners, engineers, and analysts who must collaboratively discover, structure, and query enterprise data.

Approach: DIA (Data Intelligence Agents)

DIA compresses this workflow into a system of three agents:

1. Data Interpreter — interprets raw enterprise data 2. Schema Creator — structures data into schemas 3. Query Generator — generates and executes queries

Unlike traditional text-emitting approaches, these agents treat autonomous coding agents (ACAs) as a first-class abstraction. They:

Generate, execute, validate, and repair concrete artifacts rather than emitting text
Draw on a shared memory for experience reuse
Surface outputs for review by domain experts

DIA is deployed in production for enterprise customers.

Query Generator Evaluation

The authors study the Query Generator in depth and evaluate it in fully autonomous mode across seven SQL benchmarks, covering:

Four task categories
Four SQL dialects

Result: It matches or surpasses the best published results on all seven benchmarks.

Conclusion

The results demonstrate that an architecture grounded in execution, built on ACAs and a shared memory, generalizes across the data intelligence workload—with adaptation confined to natural-language instructions.

---

*Source: arXiv:2506.14970*