Upload a candidate’s resume, choose a country, and get a full compliance brief grounded in official labor law.
Upload a candidate’s resume, choose a country, and get a full compliance brief grounded in official labor law.
How it works
Upload a resume
Drop in any PDF. The system reads name, experience, current location, and role from the document before anything reaches the compliance engine.
Choose hiring country
Pick from 8 countries and add an optional salary target. The right set of statute indexes is loaded for that jurisdiction automatically.
Get a grounded brief
Misclassification risk, mandatory benefits, notice rules, and total employer cost. Every claim includes a source document and section number.
Product thinking
A compliance tool that gives confident wrong answers is a liability, not a product. These three decisions came from that constraint first.
Citations are enforced, not suggested
The generation step can only reference passages retrieved from the indexed statute. If a claim cannot be cited, the report says so rather than producing a confident-sounding guess. The citation harness runs server-side.
Tree navigation, not similarity search
Most AI retrieval pipelines score query-chunk similarity over fragmented text. For dense legal documents with numbered articles and chapters, HireCompass navigates the table of contents directly, targeting the exact provision.
Government portals only
Every indexed document comes from an official source: legislation.gov.uk, gesetze-im-internet.de, laws-lois.justice.gc.ca. No law firm summaries, no aggregators, no paraphrased explanations of the original statute.
Architecture
Retrieval and generation are different workloads. Separating them means each service is optimized for exactly one job.
Next.js
parse-resume + /api/analyze
OpenRouterFastAPI
PageIndex tree navigation
OpenRouter8 Country JSON Indexes
Compliance Brief
Grounded, cited, streamed live
Handles the UI, resume extraction, and compliance report generation. Model calls route through OpenRouter. Results are streamed to the browser via Server-Sent Events as the model writes.
Loads all 8 country indexes into memory at startup. Receives retrieval queries from Next.js, navigates the statute tree, and returns the exact relevant passages. A silent health ping warms the Render instance the moment a user starts filling out the form.
One hierarchical JSON tree per source document, committed to the repo and loaded once at startup. No vector database, no embedding pipeline, no chunking strategy to tune. Countries with multiple statutes each get one index per document.
Routes all LLM calls across both services. The switch from direct OpenAI was deliberate: OpenAI's free tier caps gpt-4o-mini at 10K requests per day, and Saudi Arabia's index build alone hit that ceiling. OpenRouter bills per token with no daily cap, and falls back to direct OpenAI when its key is absent.
Coverage
Every index is built from a government or legislative portal. Every citation links back to the exact statute and section.
Australia
legislation.gov.au
Fair Work Act Vols 1+2 · SGA
Jordan
mol.gov.jo
Labour Code
Saudi Arabia
hrsd.gov.sa
Labour Law
UAE
uaelegislation.gov.ae
Employment Decree-Law · Cabinet Resolution
United Kingdom
legislation.gov.uk
ERA 1996 · Working Time Regulations
Germany
gesetze-im-internet.de
KSchG · BUrlG (translated)
United States
dol.gov
FLSA (federal scope)
Canada
laws-lois.justice.gc.ca
Canada Labour Code Part III
Upload a resume, pick a country, and get a grounded compliance brief in under a minute.
