feat: ship premium IT-focused vacancy radar

apps/jobs/services/relevance.py

@@ -0,0 +1,180 @@
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+
+from django.db.models import Q
+
+from apps.profiles.models import SearchProfile
+
+from .normalization import normalize_token
+
+# Deliberately title-led: employer pages regularly mention digital systems in otherwise
+# non-IT vacancies. A description hit alone must therefore never make a vacancy relevant.
+IT_TITLE_TERMS = (
+    "it",
+    "ict",
+    "informatica",
+    "software",
+    "development engineer",
+    "devops",
+    "secops",
+    "cloud",
+    "data engineer",
+    "data scientist",
+    "data steward",
+    "database",
+    "dba",
+    "business intelligence",
+    "bi analyst",
+    "system engineer",
+    "systems engineer",
+    "systeembeheer",
+    "infrastructure engineer",
+    "infrastructuur engineer",
+    "network engineer",
+    "network architect",
+    "netwerkbeheer",
+    "netwerk engineer",
+    "cybersecurity",
+    "cyber security",
+    "security engineer",
+    "security architect",
+    "security analyst",
+    "information security",
+    "informatiebeveiliging",
+    "workplace engineer",
+    "modern workplace",
+    "digital workplace",
+    "service desk",
+    "servicedesk",
+    "helpdesk",
+    "support engineer",
+    "support specialist",
+    "application lead",
+    "application engineer",
+    "application manager",
+    "application specialist",
+    "applicatiebeheer",
+    "platform engineer",
+    "platform expert",
+    "solution architect",
+    "solutions architect",
+    "technical architect",
+    "technisch architect",
+    "functional analyst",
+    "functioneel analist",
+    "business analyst",
+    "product owner",
+    "web developer",
+    "mobile developer",
+    "low-code developer",
+    "low code developer",
+    "api developer",
+    "idm developer",
+    "power platform",
+    "forgerock",
+    "embedded",
+    "c#",
+    "front end",
+    "frontend",
+    "back end",
+    "backend",
+    "full stack",
+    "fullstack",
+    "php",
+    ".net",
+    "java",
+    "python",
+    "powershell",
+    "azure",
+    "aws",
+    "microsoft 365",
+    "linux",
+    "windows server",
+    "kubernetes",
+    "docker",
+    "terraform",
+    "erp",
+    "sap",
+    "machine learning",
+    "ai engineer",
+    "artificial intelligence",
+    "image processing",
+    "computer vision",
+    "qa engineer",
+    "test automation",
+)
+
+# These phrases can contain IT vocabulary while describing commercial, recruitment,
+# or educational-design work. They therefore override positive title signals.
+NON_IT_TITLE_TERMS = (
+    "business developer",
+    "business development",
+    "it recruitment",
+    "it recruiter",
+    "learning designer",
+)
+
+IT_PROFILE_TERMS = (*IT_TITLE_TERMS, "networking", "security", "active directory", "intune")
+
+
+def _term_pattern(term: str) -> str:
+    escaped = re.escape(term).replace(r"\ ", r"[\s/_-]+")
+    return rf"(?<!\w){escaped}(?!\w)"
+
+
+IT_TITLE_PATTERN = "(?:" + "|".join(_term_pattern(term) for term in IT_TITLE_TERMS) + ")"
+NON_IT_TITLE_PATTERN = "(?:" + "|".join(_term_pattern(term) for term in NON_IT_TITLE_TERMS) + ")"
+_IT_TITLE_RE = re.compile(IT_TITLE_PATTERN, re.IGNORECASE)
+_NON_IT_TITLE_RE = re.compile(NON_IT_TITLE_PATTERN, re.IGNORECASE)
+
+
+@dataclass(frozen=True)
+class ItRelevanceAssessment:
+    relevant: bool
+    signals: tuple[str, ...]
+    reason: str
+
+
+def assess_it_relevance(title: str) -> ItRelevanceAssessment:
+    normalized_title = normalize_token(title)
+    blocked_signal = _NON_IT_TITLE_RE.search(normalized_title)
+    if blocked_signal:
+        return ItRelevanceAssessment(
+            relevant=False,
+            signals=(),
+            reason=f"Niet-technische titelcontext: {blocked_signal.group(0)}.",
+        )
+    signals = tuple(
+        term
+        for term in IT_TITLE_TERMS
+        if re.search(_term_pattern(normalize_token(term)), normalized_title, re.IGNORECASE)
+    )
+    if signals:
+        return ItRelevanceAssessment(
+            relevant=True,
+            signals=signals[:4],
+            reason="IT-signaal in functietitel: " + ", ".join(signals[:4]),
+        )
+    return ItRelevanceAssessment(
+        relevant=False,
+        signals=(),
+        reason="Geen aantoonbaar IT-signaal in de functietitel.",
+    )
+
+
+def profile_requires_it_focus(profile: SearchProfile) -> bool:
+    configured = " ".join([profile.name, *profile.desired_titles, *profile.desired_skills])
+    normalized = normalize_token(configured)
+    return any(
+        re.search(_term_pattern(normalize_token(term)), normalized, re.IGNORECASE)
+        for term in IT_PROFILE_TERMS
+    )
+
+
+def it_relevance_query(prefix: str = "") -> Q:
+    """Return the database equivalent of the conservative title-led classifier."""
+    return Q(**{f"{prefix}original_title__iregex": IT_TITLE_PATTERN}) & ~Q(
+        **{f"{prefix}original_title__iregex": NON_IT_TITLE_PATTERN}
+    )

apps/jobs/services/scoring.py

@@ -14,6 +14,7 @@ from apps.jobs.services.geocoding import resolve_cached_location
 from apps.profiles.models import SearchProfile
 
 from .normalization import normalize_token
+from .relevance import assess_it_relevance, profile_requires_it_focus
 
 
 @dataclass(frozen=True)
@@ -227,6 +228,9 @@ def _hard_exclusions(
     conflicts = sorted(excluded_skills.intersection(explicit_job_skills))
     if conflicts:
         reasons.append("Uitgesloten verplichte skill: " + ", ".join(conflicts))
+    it_relevance = assess_it_relevance(job.original_title)
+    if profile_requires_it_focus(profile) and not it_relevance.relevant:
+        reasons.append(it_relevance.reason)
     return reasons, distance.distance_confidence
 
 
@@ -292,6 +296,7 @@ def _analyze_with_ai(job: JobPosting, profile: SearchProfile) -> tuple[AiAnalysi
 def calculate_score(job: JobPosting, profile: SearchProfile) -> ScoreResult:
     distance = _distance(job, profile)
     exclusions, distance_confidence = _hard_exclusions(job, profile, distance)
+    it_relevance = assess_it_relevance(job.original_title)
     title_fit = _title_fit(job, profile)
     skill_fit, present_skills, missing_skills = _skill_fit(job, profile)
     features = job.analysis_features or {}
@@ -381,6 +386,8 @@ def calculate_score(job: JobPosting, profile: SearchProfile) -> ScoreResult:
     concerns: list[str] = []
     if title_fit >= 0.75:
         positives.append("Functietitel sluit goed aan op het zoekprofiel.")
+    if it_relevance.relevant:
+        positives.append("IT-focus bevestigd via de functietitel.")
     if present_skills:
         positives.append("Herkenbare skills: " + ", ".join(present_skills[:6]))
     if job.direct_employer and not job.recruiter:
@@ -435,6 +442,11 @@ def calculate_score(job: JobPosting, profile: SearchProfile) -> ScoreResult:
         concerns=concerns,
         hard_exclusions=exclusions,
         evidence={
+            "it_relevance": {
+                "relevant": it_relevance.relevant,
+                "signals": list(it_relevance.signals),
+                "reason": it_relevance.reason,
+            },
             "distance_km": distance.exact_distance_km,
             "distance_has_data": distance.has_distance_data,
             "distance_exact": distance.exact_distance_km is not None,