KKhang

KHANG PUBLICATION / AI ENGINEERING

I build AI systems and show the work.

My personal blog and portfolio for agent systems, applied AI projects, evaluations, failures, and lessons from real builds.

View AI projectsRead the blog
Latest articleRead article

How I Evaluated an Ecommerce Analytics Agent: Answers, SQL, Tool Use, and Safety

A public Olist case study covering answer quality, tool use, SQL safety, and the limits of local agent evaluations.

Sep 16, 2026Read article ↗

See what
I'm building

01

Local evaluation

Olist Ecommerce Analytics Agent

17/17 ADK and 12/12 agents-cli evaluations passed; not production-tested.
Details →
AnthropicAnthropicOpenAIOpenAIGoogle GeminiGoogle GeminiGrokGrokDeepSeekDeepSeekGLM / Z.aiGLM / Z.aiQwenQwenKimiKimiMiniMaxMiniMaxPythonPythonPower BIPower BIBigQueryBigQueryGoogle ADKGoogle ADKHermes AgentHermes AgentClaude CodeClaude CodeCodexCodexGitGitDockerDockerKubernetesKubernetes

How I approach a build

Start with a question. Make something small. Test the assumptions, then share what the process taught me.

More about me
Understand the right problem, test what matters, and learn from the live system.