I sit at the intersection of building and writing.

I’m Hande Kafkas. By day I’m an AI engineer — I design and ship the systems that turn language models into products people can actually rely on. By all the other hours, I’m a technical writer trying to describe what I’m seeing while it’s still warm.

The two halves feed each other. Writing forces me to slow down and name what I’m doing; building keeps the writing honest. The work I care about is the part that doesn’t fit in a demo: evals, retrieval, planning, the long tail of failure modes that only show up under real load.

What I work on

• Agent runtimes — making the plan a first-class artifact, not an emergent property of attention.
• Retrieval & evals — catching the silent regressions that ship to production unnoticed.
• Tools for technical writers — small, voice-preserving writing aids, not generic AI prose generators.

Working with me

I take on a small number of consulting engagements per quarter, usually with teams who already have an LLM product in production and are running into the parts that aren’t the model. If that sounds like you, say hi on X or find me on LinkedIn.