JobGen

An AI-assisted job platform that aggregates remote roles, matches them to candidate profiles, and supports CV improvement workflows.

View source Open live project

Overview

The system, not just the screen.

JobGen combines a Go backend with a React and TypeScript client. The backend coordinates job ingestion, matching, CV processing, AI calls, authentication, and storage behind REST APIs.

Problem

Remote roles are fragmented across sources, and early-career candidates need a clearer way to understand how their profile maps to each opportunity.

My work

Built the backend service structure and web product around job aggregation, profile-based matching, CV workflows, and AI-service integration.

Architecture

How data moves through the system.

01React + TypeScriptCV and job-search experience
02Go REST APIAuth, orchestration, validation
03Domain servicesAggregation, matching, CV workflows
04Data & integrationsMongoDB, Redis, MinIO, Gemini, job sources

Engineering

Decisions and constraints.

Selected decisions

01
Run independent job-source adapters concurrently with goroutines and sync.WaitGroup, then collect failures without discarding successful sources.
02
Keep source-specific scraping behind interfaces so sources can be registered or removed without rewriting the aggregation flow.
03
Separate deterministic skill, experience, and location scoring from AI-assisted CV features.
04
Use Redis for cache/session concerns and MongoDB for application data; keep object storage behind MinIO-compatible APIs.

Challenges

Normalizing inconsistent job descriptions and skill labels across multiple external sources.
Bounding slow external work with context deadlines and keeping partial source failures visible.
Balancing explainable matching rules with AI-assisted workflows.

representative_flow.go

group, failures := sync.WaitGroup{}, make(chan error, len(sources))
for name, source := range sources {
    group.Add(1)
    go func() {
        defer group.Done()
        if err := aggregate(ctx, source); err != nil {
            failures <- fmt.Errorf("%s: %w", name, err)
        }
    }()
}

Stack & evidence

What it was built with.

Go
Gin
MongoDB
Redis
MinIO
React
TypeScript
Docker
GitHub Actions