OPENAI SEOUL

Application Packet

Doeon Kim — Solutions Architect / Data Platform Engineer

OpenAI Alignment

stage-pilot — Tool-calling reliability runtime for AI SDK applications (npm: @ai-sdk-tool/parser). 1,720 tests ensuring robust structured output from OpenAI and other models.
tool-call-finetune-lab — QLoRA fine-tuning pipeline for tool-calling improvement, evaluated on the BFCL benchmark — directly advancing function-calling reliability.
Nexus-Hive — LangGraph multi-agent NL-to-SQL system demonstrating complex agent orchestration and tool use patterns.
enterprise-llm-adoption-kit — Production LLM governance with evals, PII redaction, RBAC — helping enterprises adopt AI safely.

Technical Depth

Tool-calling: Published npm reliability runtime, streaming parsers, structured output validation
Fine-tuning: QLoRA, PEFT, Hugging Face Transformers, BFCL benchmark evaluation
Agent systems: LangGraph orchestration, multi-agent collaboration, governed tool use
Multimodal: Vision + text incident analysis with Gemini (AegisOps, live demo)
Safety: PII redaction, RBAC, eval frameworks, responsible AI deployment patterns

Why OpenAI Seoul

I ship production AI infrastructure that makes LLMs more reliable and accessible. My tool-calling runtime and fine-tuning work directly improve the developer experience OpenAI is building for the world.

Credentials

Microsoft AI-900
SnowPro Associate & Databricks Platform Architect
Datadog Observability Fundamentals
Trilingual: Korean (native) / Japanese / English — Seoul based

stage-pilot Fine-Tune Lab