How job titles rise and fade

An engineer who works inside a customer's environment to ship the software they need, part builder and part consultant. Palantir made the title known; its first Hacker News Who is hiring sighting is a September 2011 post from Addepar. AI labs and AI startups adopted it in 2024 and 2025, and it took off in late 2025: it is now in about 5% of Who is hiring posts, up from almost none two years earlier.

Someone who designs what a language model sees: the retrieval, memory, tool results and instructions that fill its context window. The term spread in mid 2025 as the successor to prompt engineering. It is still rare as a formal job title; Titles tracks every posting that uses it, including variants such as Context Architect.

The ones gaining share fastest or appearing for the first time: forward deployed engineer, AI engineer, member of technical staff, applied AI engineer and agent engineer on Hacker News, plus new titles on company job boards such as deployment strategist, applied AI architect, AI deployment engineer and legal engineer. The emerging page lists them with first sightings.

An ML engineer trains, ships and maintains machine learning models. An AI engineer mostly builds products on top of models someone else trained, calling them through APIs, with retrieval, agents and evals. In Hacker News Who is hiring threads AI engineer has been mentioned more often than ML engineer every month since December 2025.

Two public sources. Every monthly Hacker News "Ask HN: Who is hiring?" thread since April 2011, read through the free Algolia API: a title's share is the percent of that month's postings that mention it. And the public job boards (Greenhouse, Ashby, Lever) of about 130 data, AI and tech employers, read weekly: open roles by title, and salary ranges where a posting states one.

Only from postings that state a pay range themselves (many US postings must). The median is the middle of the midpoints of those ranges, in US dollars per year. Postings without a range are left out, so a title with few stated ranges has a rough median.

Every title and company page has Suggest a correction, and the listing pages have Suggest a missing title. Each suggestion is checked before anything changes.