About

"Sometimes you have to rise above the clouds to look down —
only then does the full structure and context reveal itself."

JUNHO LEE / 이준호 : Data & AI Platform Architect · PM

BATUR MOUNTAIN
BALI · INDONESIA

PROFILE / 01Junho Lee · 이준호

Designing the semantic structure between language and data.

Started as a Web/ERP developer, then progressed through DW/BI engineer, Technical Lead, and Consulting Division Head. Led large-scale DW cloud migrations and operated next-generation data platform projects as a multi-vendor PMO.

Grew a small consulting team to division scale, with revenue growing several times over. Not just an engineer — handled hiring, training, Presales, and C-level seminars.

Now working at the intersection of data and AI. Building an LLM-based BI Agent, I hit the real-world limits of NL2SQL firsthand, and that process led to the conviction that ontology is necessary. In the first half of 2026, I built DataNexus as a personal project around that idea and documented it as a build log.

Years of work with data. Watched how good questions and well-designed models change decisions.
Now designing the semantic structure between language and data. Designing the semantic structure between language and data.

FIELD NOTES · 04 / 2026

BATUR MOUNTAIN BALI · INDONESIA

CHAPTER 01 : Career Arc

Working with data across two decades,
always running into the same problem.

PHASE 042024 →NOW

LLM × Ontology

Building an LLM-based BI Agent, I saw exactly where NL2SQL breaks in real environments. Abbreviated tables like T_CUST_MST and "net revenue" meaning different things across departments — problems that DDL alone cannot solve. That's why I built DataNexus, an ontology-based data platform, as a personal project in the first half of 2026.

OntologyNL2SQLGraphRAGLangChainMCP

PHASE 032019 →LEAD

Consulting Division Head · Building an Org

Grew a small consulting team to division scale, with revenue growing several times over. Took on hiring, training, Presales, and C-level seminars — learned what it means for an engineer to run an organization.

HiringPresalesPMOP&L

PHASE 022014 →SCALE

Enterprise DW · Cloud Migration

Retail, telecom, manufacturing, construction — large-scale DW projects across industries. Tech Lead on on-prem to cloud migration, and multi-vendor PMO on next-generation data platform builds.

SynapseBigQueryRedshiftYellowbrickADF

PHASE 012005 →FOUNDATIONS

From Web/ERP Developer to the Data Path

From handling field transactions, shifted to designing the data flow itself. The era when I first felt "a single query can change a department's KPI."

C#OracleSAPSSIS

"Everyone can be
an analyst."

: DataNexus Principle

CHAPTER 02 : DataNexus

The problem of working with data
in natural language.

Enterprise data analytics always runs into the same wall. "Explore internal data through natural language" is easy to say — but in practice, table names look like T_CUST_MST, and "net revenue" carries different calculation logic across departments. LLMs cannot understand business context from DDL alone.

DataNexus tackled this by combining an ontology-native NL2SQL engine, GraphRAG, and a Data Catalog. An open-source composite architecture that handled both unstructured documents and structured databases together.

From February to July 2026, this blog documented the process, including the dead ends. Architecture decisions, reasons behind technology choices, failed experiments alongside successful ones.

CHAPTER 03 : Stack

Tools validated
in the field.

ai · ml.zsh

jhlee@nexus ~ cat ai-ml.txt

Ontology · LLM RAG · NL2SQL
LangChain · MCP · GraphRAG
Multi-agent system design
Prompt orchestration · evaluation

dw · platform.zsh

jhlee@nexus ~ ls dw/

Azure Synapse · BigQuery · Redshift
PostgreSQL · Oracle · Yellowbrick
Palantir Foundry
Cloud migration · multi-tenant DW

bi.zsh

jhlee@nexus ~ bi --list

Power BI · Tableau
MicroStrategy · Qlik Sense
Looker · Apache Superset
Semantic layer design

etl · elt.zsh

jhlee@nexus ~ find . -name 'pipelines'

Azure Data Factory · Databricks
SAP Data Services · Informatica
IBM DataStage · SSIS
Medallion (Bronze · Silver · Gold)

cloud.zsh

jhlee@nexus ~ cloud --providers

Azure · Synapse, ADF, ML
AWS · Redshift, S3, Glue
GCP · BigQuery, Gemini
IaC · CI/CD for data

graph · catalog.zsh

jhlee@nexus ~ graph --status

DataHub
Neo4j · DozerDB
ApeRAG
JSON-LD · schema.org for GEO

CHAPTER 04 : Now

What I'm focused on
this quarter.

Building

Curation Auto-Publishing Pipeline

Refining the pipeline that takes links from a sheet through scraping, summarizing, and style correction into daily curations.

Writing

Claude Dev Digest

Explaining posts from Anthropic's developer blog in plain terms, with notes from the perspective of building data and AI agents.

Reading

Kimball, Revisited / Foundry Cases

Re-examining whether DW modeling from the Kimball era still holds in the cloud age.

CHAPTER 05 : Contact

Write long,
talk short.

AI agents, NL2SQL, ontology, DW consulting:
Feel free to reach out on any topic. Fastest on LinkedIn.

// 01 · LINKEDIN linkedin.com/in/leejuno → // 02 · BLOG datanexus-kr.github.io → // 03 · CURATIONS Daily Curations →