SKIP TO CONTENT

// METHODOLOGY

How the State of Agent Engineering dataset was built

How the 119-posting, 32-employer snapshot was assembled. The download contains extracted fields, source links and collection metadata. Raw responses, the collection script and the probe log are not included, so this is a method summary rather than a complete reproduction package. Postings may have changed or closed.

1. Sources, and why only these three

Data came from exactly three public, unauthenticated ATS job-board APIs, chosen because each operator documents a public job-listing endpoint:

The collection excluded LinkedIn, Indeed, Wellfound/AngelList, YC Work-at-a-Startup, Google for Jobs and republished datasets from those platforms. It used extracted fields and source links for reference analysis. Public API availability does not remove an operator's terms or settle reuse rights for every purpose.

2. Collection

A hand-curated seed list of roughly 170 mostly US, venture-backed, AI-native and infrastructure companies, extending the seed list used for the FDE study with AI-agent-tooling and applied-AI companies (model labs, agent-framework vendors, evals platforms, LLM-infra startups). Each company's board slug was resolved from its public careers URL, then its one public JSON endpoint was fetched once per ATS, rate-limited to no more than one request per second, with an identified User-Agent. No login, no paywall bypass, no IP rotation. 504 requests were made against 507 candidate (company, ATS) pairs; 102 boards returned a non-empty public job list. The metadata names the probe log, but that log is not part of the public download.

COLLECTION PIPELINE
Flow diagram: 7 stepsFlow diagram: 7 steps. Seed list, then Resolve board slugs. Resolve board slugs, then Fetch public ATS JSON. Fetch public ATS JSON, then 102 boards returned jobs. 102 boards returned jobs, then Title gate. Title gate, then Keyword coding rubric. Keyword coding rubric, then 119 rows, 32 employers.1Seed list~170 mostly US,venture-backed, AI-native andinfra companies, curated byhand.2Resolve board slugsFrom each company's publiccareers URL. No login, nopaywall bypass, no IProtation.3Fetch public ATS JSON504 requests across 507candidate (company, ATS)pairs, <=1 req/sec, identifiedUser-Agent.4102 boards returned jobsThe rest were empty, moved, ornot on Greenhouse/Lever/Ashby.5Title gateAgent / AI-engineer titlefamily only. ML Engineer andData Scientist are out ofscope.6Keyword coding rubricAgent-framework, evals andautonomy signals coded fromkeyword matches; the raw textis not included.7119 rows, 32 employersExtracted fields andlink-backs only. No verbatimposting prose.
The published metadata records the collection scope. The probe log and raw responses are not included in the download, so request counts cannot be independently reproduced from it.

3. Inclusion and exclusion

Included if the posting was live and its title matched the agent/AI-engineer-family lexicon: AI Engineer, Applied AI Engineer, Agent Engineer, Forward-Deployed AI Engineer, Prompt Engineer, or an evals/AI-safety specialist title.

Excluded: generic Software Engineer or ML Engineer postings with no AI-agent-family title match, pure data-science/research-scientist roles, internships, expired postings, and duplicate cross-posts. Deduplication is keyed on canonical employer name (not raw board slug) plus normalized title and location, so a company listing the same role under two slug aliases or on two ATS platforms is counted once, not twice.

4. The agent-signal coding rubric

Unlike the FDE study's five-axis authenticity score, this study codes four independent binary signals per posting, each keyed to a vocabulary match. The published table retains the resulting fields but not the source-text spans:

SignalTriggered by
Agent framework / agentic patternLangChain, LangGraph, CrewAI, AutoGen, AutoGPT, LlamaIndex, “agentic,” “multi-agent,” tool use, function calling, orchestration
Production codinglanguage/stack requirements consistent with shipping production software, not aptitude-only screening
Evals / red-teaming / guardrails“evals,” “red-team,” “guardrails,” safety-review language
Autonomy / human-in-the-loop“autonomy,” “human-in-the-loop,” “escalation,” oversight language

LLM-stack mentions (RAG, OpenAI API, Anthropic/Claude, vector databases) and programming languages are coded the same way, as independent keyword matches, not a composite score. The keyword rules are not human-adjudicated for every row and can miss context, so treat them as coded signals rather than verified job requirements.

5. What gets published (and what doesn't)

Published: aggregate statistics, a per-posting fact table of extracted attributes, and a link back to each live posting. Notpublished: verbatim job-description prose (that is the employer's copyright), and anything from a prohibited source. Source links identify where the fields came from; they are not a blanket license for other uses.

6. Limitations

This is agent-engineering hiring among the AI-native and infrastructure companies in the seed list, on public ATS boards, not the whole market. The 119 postings and 32 employers form a convenience sample, not a representative survey. A larger sample alone would not remove the selection bias. It is a point-in-time snapshot (open roles, not hires), so trend claims need a second edition. This corpus is never blended with the 519-posting FDE dataset at /state-of-fde- different title family, different collection round. Full limitations are listed on the report.

▸ BACK TO THE REPORTDOWNLOAD THE DATASET