I will design and build a safe llm agent with guardrails

T
tonylim8009
T
tonylim8009
JongChul L

About this gig

I design and build LLM agents for customer support, internal knowledge and regulated domains such as healthcare and finance, where wrong answers are a real cost.


What I design:

- Intent classification and safety gates that run before the LLM

- Retrieval (RAG) with category isolation. If no source is found, the agent does not answer

- Explicit dialog states, with handoff to a human agent

- Evaluation: regression test set, pass/fail gates in CI, coverage report


Models: API models (GPT, Claude) or on-prem models served with vLLM.


Packages:

- Basic: review of your current design, written report with fixes

- Standard: full design document and evaluation plan with a test set

- Premium: design, working MVP, regression suite and evaluation report


How we work: in writing. Send your use case, data sources and model choice before ordering. Everything is delivered as files your team can keep.


Not included: model fine-tuning, production hosting.


Get to know JongChul L

JongChul L

Head of AI RD Center

  • FromSouth Korea
  • Member sinceOct 2026
  • Languages

    Korean
I have spent about 30 years building telecom software, including network protocols, VoIP, IMS, and SIP systems. Currently, I lead the AI R&D Center at UANGEL, where I design carrier-grade systems that integrate LLMs and speech AI under strict telecom constraints. I specialize in voice AI, real-time communication, and cloud-native MLOps.

My Portfolio