Skip to content
LANGOLF DIGITAL
All work01 / 15
AI2025

Job Hunter AI

Job discovery, matching, and application intelligence platform

An AI-assisted job discovery and intelligence platform: discover and normalize listings, match candidates to roles with deterministic scoring, analyze opportunities, track applications, and keep follow-ups moving.

Status
Functional Prototype
Category
AI
Year
2025
Primary stack
Next.jsTypeScriptSupabase
Demo
ENTER THE BUILD
Source
GitHub
Job Hunter AI interface visualization

Project visualization — a designed representation, not a literal screenshot

01Overview

What this is

An AI-assisted job discovery and intelligence platform: discover and normalize listings, match candidates to roles with deterministic scoring, analyze opportunities, track applications, and keep follow-ups moving.

Explores

  • Job discovery from multiple sources
  • Job normalization and deduplication
  • Candidate-to-role matching
  • Deterministic scoring and ranking
  • AI-assisted job analysis
  • Resume management
  • Application tracking
  • Scheduled discovery
  • Notifications
  • Supabase / PostgreSQL storage
  • Supabase Edge Functions
  • Automated testing

02Explore the build

The demo and the build insight

The demo

External application

ENTER THE BUILD

A functional prototype of the job-discovery, matching, application, and follow-up workspace. Uses a verified, publicly available deployment.

ENTER THE BUILD

Build insight

01 / 04
Current build
Functional Prototype
Architecture and features below are the real build.
Demo
External
Launch control above opens the verified public build.

Build insight — real project data, not a live application.

The app separates a deterministic matching core from the AI analysis layer. Raw listings flow through normalization and deduplication into PostgreSQL; the scoring engine reads the candidate profile and emits interpretable scores; AI analysis runs as a structured post-processing step on a shortlist rather than on every raw row.

03Problem · opportunity

What it sets out to solve

Job searching is fragmented across job boards, spreadsheets, resumes, email, and follow-up reminders. Candidates struggle to know which roles are genuinely worth their time and which steps still need a follow-up.

04Concept

The core idea

A single workspace that turns raw job listings into normalized, scored, and trackable opportunities — pairing deterministic matching logic with AI analysis so the candidate understands why a match is strong or weak.

05What I built

What's actually in the code

A functional prototype with a working job-discovery and application-tracking loop powered by Supabase (PostgreSQL) and Supabase Edge Functions. The current implementation includes deterministic matching and ranking, an AI job-analysis step, resume management, scheduled discovery, and automated tests around the core scoring logic.

Approach

Built the system as a modular pipeline: ingest listings, normalize and deduplicate them into a PostgreSQL schema, score roles against a candidate profile using deterministic rules, and layer on AI analysis for job descriptions and resume tailoring. Discovery runs on a schedule, with notifications for new matches.

06Key features

What's implemented today

  • Pipeline that ingests raw listings, normalizes fields, and removes duplicates before they enter the database
  • Deterministic candidate scoring so ranking is explainable rather than a black box
  • AI step that reads a specific role and returns a structured analysis of what the position is really asking for
  • Application tracker with statuses and follow-up reminders
  • Scheduled discovery jobs that pull new opportunities and trigger notifications
  • Automated tests covering the normalization, matching, and scoring modules

07Architecture

How it's structured

architecture.md

The app separates a deterministic matching core from the AI analysis layer. Raw listings flow through normalization and deduplication into PostgreSQL; the scoring engine reads the candidate profile and emits interpretable scores; AI analysis runs as a structured post-processing step on a shortlist rather than on every raw row.

Design notes

Resumes and personal data are treated as sensitive. The matching/ranking logic is written as pure functions so it can be verified with automated tests independent of the UI and the network.

08Technology

The stack

  • Next.js
  • TypeScript
  • Supabase
  • PostgreSQL
  • Edge Functions
  • Node.js
  • Tailwind CSS

09Challenges · lessons

Where it got hard, and what it taught me

Challenges

  • Deduplicating the same role across multiple job boards without losing source-specific details
  • Designing scoring rules that are honest — assignment-weighted, transparent, and free of fabricated precision
  • Keeping AI analysis bounded to a structured schema so output remains stable and testable

Learnings

  • Deterministic rules and AI analysis complement each other: rules guarantee explainability, while AI adds nuance a rules engine cannot express
  • A pipeline-shaped architecture makes it easy to test each stage in isolation
  • Notification and scheduling design matters as much as the matching itself for real-world usefulness

10Current status

Where this project sits today

Functional Prototype2025

Core features are engineered and demonstrable. Not yet production-hardened or deployed to real users.

Demonstrated

  • Functional prototype with a working match-score flow and application tracker
  • Automated tests run against the normalization, deduplication, and scoring modules

11What comes next

Where this goes from here

  1. 01Persist verified candidate outcomes — such as applications actually sent — through a controlled, user-authored flow
  2. 02Add broader discovery source adapters
  3. 03Harden the Edge Functions and add CI coverage across the full pipeline