TAXONOMY.md

SkillJev Taxonomy v0.1 — Backend / Fullstack (IT/Engineering)

10 competency groups (a JD maps to 6–10 of them = radar axes). Each group has stable group_id; skills are leaves (skill_id, lowercase snake_case). The skill head predicts group_id.

group_id	Name	Example skills (leaves)
lang_core	Programming Languages & Core CS	java, go, python, node_ts, kotlin, rust, concurrency, data_structures
api_design	API & Service Design	rest, grpc, graphql, api_versioning, idempotency, openapi
data_storage	Data Storage & Modeling	postgresql, mysql, mongodb, redis, schema_design, query_optimization, migrations
system_design	Distributed Systems & Architecture	microservices, event_driven, kafka, caching_strategy, scalability, consistency
cloud_devops	Cloud, Containers & Delivery	aws, gcp, docker, kubernetes, terraform, ci_cd
reliability_obs	Reliability & Observability	monitoring, tracing, slo_sli, incident_response, performance_tuning
security	Application & Infra Security	authn_authz, oauth_oidc, owasp, secrets_mgmt, encryption
testing_quality	Testing & Code Quality	unit_testing, integration_testing, tdd, code_review, static_analysis
frontend	Frontend (fullstack roles)	react, typescript_fe, state_mgmt, web_performance, accessibility
leadership	Technical Leadership & Collaboration	mentoring, tech_design_docs, cross_team, ownership, hiring

Rules

• One requirement → exactly one primary group_id (secondary groups allowed in meta.secondary_groups, not trained in v1).
• Unmapped skills → other (kept out of v1 training; logged for taxonomy growth).
• Synonyms normalized in taxonomy/aliases (e.g. "k8s"→kubernetes, "Postgres"→postgresql, "AWS ECS"→aws+docker).
• Taxonomy is versioned (taxonomy_version: "0.1"); every sample stores it.