Nick Khami

Khami earned a computer-science BS at the University of Texas at Austin in 2022, with a machine-learning concentration. He founded Trieve, a search API company, and joined Mintlify after its acquisition. He moved to inference work at Morph in August 2026. His inference writing separates routing, model serving and cache reuse into different engineering responsibilities.[1][2][5]

Let a software agent complete a telephone step

call4me gives an agent MCP tools for calling a business, navigating menus and returning an outcome and transcript. The service connects a telephony provider with a realtime voice model, with separate tools for ending the call, asking the user and pressing digits. A call can therefore bring a task back to the person or agent when information is missing.[3][4]

Treat telephone menus as changing state

The repository distinguishes a fresh menu prompt from transcript timestamps and remembers accepted keypad submissions. Recovery waits for pending tools and avoids interrupting a hold or a person answering. This addresses the practical difference between generating plausible speech and completing an interaction with a live telephone system.[4]

His inference skills map describes routing and scheduling, the runtime that executes model requests, and caches that reuse earlier work. The layers interact but solve different problems. This gives a web developer entering inference a way to identify which service or bottleneck they need to understand before changing the whole deployment.[5]