sahityabiswas/job_nexus

Purpose and Entry Points

JOB NEXUS is a career recommendation system that stores jobs, skills, and courses in Neo4j, then recommends jobs from a user skill profile. Runtime entry points are the FastAPI service in src/api.py, the standard ingestion command in src/ingest_simple.py, and a scalable Spark ingestion alternative in src/ingest_spark.py. The browser SPA is static/index.html.

API and SPA Boundary

src/api.py serves the SPA and exposes REST endpoints for skill search/normalization, filter metadata, recommendations, and graph data. static/index.html owns interactive profile entry, filtering, recommendation display, graph visualization, sharing, and PDF export. The frontend communicates only through the FastAPI API boundary; matching and graph queries remain server-side.

Recommendation Pipeline

src/recommender.py is the ranking engine. It accepts normalized user skills plus proficiency and filtering/sort inputs, retrieves job requirements from Neo4j, and combines graph-overlap information with TF-IDF/cosine similarity. Its outputs include match scores and matched/missing skill breakdowns used by the recommendation UI and learning simulation.

Skill Canonicalization

src/normalize.py centralizes alias mapping and canonical skill lookup. Both ingestion and request-time profile handling should pass through this layer so variants such as misspellings or alternate spellings resolve to one Skill node identity. This is a key consistency boundary between raw external records/user input and the graph schema.

Graph Persistence and Query Layer

src/db_neo4j.py manages the Neo4j driver/connection pool and isolates AuraDB credentials and database access from API, matching, and ingestion code. The graph’s architectural entities are Job, Skill, and Course, connected by REQUIRES and TEACHES relationships. Neo4j supports both recommendation retrieval and the graph payload rendered in the client.

Ingestion Paths

src/ingest_simple.py is the local/default loader, likely consuming repository seed datasets such as data/courses.csv and data/users.csv before creating normalized nodes and relationships. src/ingest_spark.py is the distributed path: it fetches public job/course data, partitions processing with Spark, performs NLP normalization, and writes graph partitions concurrently to Neo4j. spaCy is a material dependency for skill extraction/normalization.

Runtime and Deployment

The backend is Python/FastAPI with scikit-learn vector matching, spaCy NLP, and Neo4j AuraDB as the external graph service. requirements.txt defines Python runtime dependencies. Render deployment is represented by Procfile and render.yaml; connection configuration is supplied through environment variables or the template in config.example.json.