05 / Étude de cas

Moteur ATS Intelligence

Une couche d’acquisition multi-ATS avec normalisation canonique, détection des changements et signaux de recrutement.

Beta · Source reviewed · Selected tests passing
Moteur ATS Intelligence
Contexte

Problème et valeur

The project supports recurring hiring intelligence rather than a one-off export. It is structured as a reusable Python package with database schemas, orchestration scripts, observability hooks and dedicated modules for rate limiting, proxy health, normalization, revisions and hiring velocity.

Every ATS represents similar concepts differently. Titles include requisition codes, locations combine city and work-mode signals, contracts and salaries use inconsistent formats, and boards change independently. Without a canonical model and snapshot logic, the result is duplicate data rather than intelligence.

Ingénierie

Architecture et résultat

The package targets Python 3.11+ and uses asyncio, httpx, psycopg, Selectolax and Prometheus instrumentation. Source clients are registered behind shared orchestration. Separate modules handle normalization, delta computation, revisions and HVI classification. The pyproject classifies the project as Beta, which is more accurate than unsupported claims of absolute production readiness.

The engine creates a reusable acquisition layer for job-market monitoring. It reduces downstream complexity by presenting one schema across multiple ATS platforms and separating current state from meaningful change. The HVI layer turns job lifecycle data into a company-level acceleration signal.

10 adapter source files identified
24 test modules covering infrastructure and platform clients
Captured selected run: 68 passed, 19 warnings in 1.24 seconds
PostgreSQL schemas for core records, revisions, cache and optimized indexes
CLI scripts for sweeps, exports, HVI backfill and classifier training