Personal website: https://soominmyung.com
- Multi-Agent Workflows: Designing self-correcting, state-based LLM orchestration using LangGraph to automate complex reasoning and operational tasks.
- LLM Observability: Integrating LangSmith for full telemetry tracing, evaluating AI outputs, and mitigating hallucination risks in production.
- Production-Grade RAG: Implementing high-precision retrieval using ChromaDB with advanced metadata filtering and automated vector ingestion.
- Serverless Architecture: Deploying high-performance containerized microservices to GCP Cloud Run using Docker.
- CI/CD Automation: Building robust deployment pipelines via GitHub Actions for zero-downtime updates and reliable continuous integration.
- High-Performance APIs: Building asynchronous backends with FastAPI featuring Server-Sent Events (SSE) for real-time progress streaming.
- Frontend Integration: Seamlessly connecting AI backends to modern UI platforms using React and TypeScript.
- Data Pipelines: Engineering ETL/ELT pipelines for massive-scale historical records (41M+ rows) using PySpark and Parquet.
- Forecasting & Analytics: Translating complex business logic into predictive inventory and anomaly detection models.
- Frameworks: LangGraph · LangChain · MCP (Model Context Protocol)
- LLMOps: LangSmith · Guardrailed Prompting · AI Output Evaluation
- Vector DB: ChromaDB · FAISS
- Languages: Python · TypeScript · SQL
- Backend: FastAPI · Asynchronous Programming · SSE
- Cloud & DevOps: GCP (Cloud Run, BigQuery) · Docker · GitHub Actions (CI/CD)
- Big Data: PySpark · Parquet · SQL Server (T-SQL)
- ML Modeling: PyTorch · Transformers · Time-series Forecasting (ARIMA/Prophet)
- Frontend: React · Framer Custom Code Components