Building AI agents that actually work.

Currently @ Asleep — shipping AI-powered sleep products to real users.

B2C & D2C products — not just demos, actual products people use

• Sleep analysis apps in production
• SleepHub: device-to-customer, AI meets hardware
• Voice agents that talk to real customers

Agent infrastructure — making AI agents effective at scale

• Internal skill libraries that agents can actually use
• Multi-agent orchestration for production workloads
• End-to-end: LLM pipelines → backend → cloud infra

• Ship > polish — learn from real users, not hypotheticals
• Agents as teammates — not just tools, actual collaborators
• Infra that disappears — good systems feel invisible

• Personal AI agents (yes, I have one)
• A2A protocols and multi-agent collaboration
• The gap between agent demos and agent products

• Twitter: @hankyeolkyung
• Company: Asleep

transhumanist, human