Data Infra / Platform Software Engineer

Specializing in large-scale data platforms, streaming and batch pipelines, lakehouse / warehouse systems, and context-aware GenAI agents for analytics engineering. I work across distributed systems, data infrastructure, and agentic workflows that turn messy metadata, warehouse context, and operational rules into reliable self-service data products.

zyusong614@gmail.com · LinkedIn

Building practical systems for analytics agents, metadata-aware SQL generation, and reproducible data lakehouse pipelines.

• Context-Aware Analytics Agent (CA3):

  • Created a local-first analytics agent that uses project metadata, warehouse schemas, docs, rules, tools, and chat context before generating SQL or insights.
  • Built a controlled tool-calling loop for table search, metadata inspection, read-only SQL execution, chart rendering, and result-grounded responses.
  • Designed a project context builder with multi-database connectors, MCP / skills integration, versioned filesystem context, eval cases, and a SvelteKit chat UI.

• RedLake:

  • Built a reproducible GCP data lakehouse project for Reddit data ingestion, anonymization, sentiment processing, and BigQuery analytics.
  • Implemented a dbt transformation layer with staging, intermediate, and mart models plus FAIR-style documentation and lineage.
  • Used Cloud Functions, GCS, BigQuery, dbt, Cloud Build, and Looker Studio to demonstrate an end-to-end analytics pipeline.

• Data platform agent systems that generate configs, Airflow DAGs, DQ rules, and Git-ready outputs from governed metadata.
• Context-grounded NL-to-SQL and NL-to-insights workflows with transparent reasoning, safe execution, and measurable reliability.
• Production data infrastructure for low-latency serving, PB-scale storage, streaming reliability, and operationally durable ETL / ELT.