HyeongWook Lee
About
AI engineer at LG Innotek who turns ambiguous business problems into AI systems that ship — and translates how they work back to the people who depend on them. I design and build agentic systems in production (multi-agent orchestration, RAG, full-stack apps), then sit with non-technical stakeholders to map what the technology can and can't do. The kind of engineer who can both build the solution and explain why it's the right one.
Tech Stack
Experience
LG Innotek
Gen AI Engineer
Enterprise AI Platform
Delivered an enterprise AI platform now in weekly active use by 20+ executives and team leads. It auto-ingests SharePoint documents on a 2-hour refresh into a multi-stage RAG pipeline, and lets users spin up their own agents for Q&A and research over Korean enterprise documents — putting organizational knowledge directly in non-technical hands.
Conversational Analytics (Text-to-SQL)
Built a text-to-SQL layer (DuckDB) so non-technical users could query Excel-based business data in plain language — turning data that was locked in spreadsheets into self-service analytics anyone could use.
Defect-Analysis Automation
Built a pipeline that turns raw production logs into standardized report drafts (charts + tables) and surfaces likely problem areas to engineers; the generated reports feed back into the platform's knowledge base.
AWS Migration
Partnered with LG CNS engineers to design the AWS migration architecture and cost model for transitioning the on-premise platform — owning the technical and cost trade-offs that made the move viable.
Customer Excellence Award · 1st Place — DX Division, LG Innotek — enterprise LLM platform
LG Bootcamp Innovation Award · 2nd Place — LG — emotion-detecting AI assistant
LG Innotek
Software Engineer Intern
Developed a CNN-based defect image detection system for manufacturing quality control.
Projects
The problem it solves — crude procurement still rides on a few experts' intuition. Crude Compass fuses public geopolitical and market signals into AI-written daily briefs a manager just approves or rejects, before prices move.
- —Agent Bricks multi-agent supervisor routes natural-language questions across Genie, a RAG knowledge assistant, and a recommendation agent — streaming the live sub-agent trace
- —Unity Catalog Bronze → Silver → Gold pipeline fuses 6 open data sources into a daily 0–100 procurement-pressure score
- —Deployed on Databricks Apps with Lakebase (Postgres) OLTP and Lakeflow scheduled jobs
The problem it solves — raw real-estate data becomes an interactive, AI-guided decision tool: a hands-on demo of how Cortex turns a warehouse into a product.
- —Snowflake Cortex AI advisor with 3-layer price model on 453 Seoul apartments
- —Snowpark UDF batch processing — 54,360 calculations compressed to 12 calls
- —Cortex Analyst Text-to-SQL on 8-table semantic model with real marketplace data
Certifications
Databricks Certified Data Engineer Associate
Databricks
Databricks Certified Generative AI Engineer Associate
Databricks
AWS Certified Solutions Architect – Associate
Amazon Web Services
Education
Penn State University
BS Computer Science · Minor in Mathematics
Contact
© 2026 HyeongWook Lee