01
Next.js 14 App Router for dashboard, jobs, and profile routes
WorkJob Radar AI
HR Tech · AI-assisted job search2026
Job Radar AI
A Next.js job-matching product that stores listings, scores them against a candidate profile, and supports apply-profile tailoring plus Greenhouse auto-apply workflows.
- Project
- Product build
- Year
- 2026
- Focus
- HR Tech · AI-assisted job search
Overview
What this project set out to solve
Job Radar AI helps candidates move from noisy listings to clearer priorities. The public MVP stores software job listings in MongoDB, scores them against a user profile, and exposes a dashboard, jobs list, and editable profile — with OpenAI-assisted apply tailoring and Playwright-based auto-apply for supported Greenhouse URLs.
Our role: Product design and full-stack delivery — scoring model, MongoDB schemas, Next.js UI, apply-profile selection, OpenAI tailoring hooks, and Playwright auto-apply workers.
Business challenge
The operational problem we designed around
Job seekers face noisy listings and weak match context, which makes it hard to prioritize applications that are actually a fit. Manual apply work across many boards also burns time without a clear ranking of where to focus.
Help candidates see scored matches against their profile, manage applications in one place, and optionally tailor resumes/cover letters and queue eligible Greenhouse applications automatically.
Hard parts of the problem
- 01Keeping scores explainable without drowning the feed in detail
- 02Classifying apply URLs accurately so auto-apply only runs on supported flows
- 03Balancing automation with candidate control — queue rules must stay inspectable
Solution
How the product solves it
We built a Next.js 14 App Router product with TypeScript, Tailwind, and MongoDB/Mongoose. Heuristic scoring returns a 0–100 score plus human-readable reasons. Apply profiles hold multiple resume personas; OpenAI supports tailoring; Playwright workers handle auto-apply for eligible Greenhouse jobs.
- 01Dashboard with totals for jobs, new listings, high-match roles, and saved jobs
- 02Jobs table with scores, status, and external links — responsive cards on mobile
- 03Editable candidate profile (roles, skills, locations, work modes, seniority)
- 04Apply profiles for multiple resume/cover personas with best-profile selection per job
- 05OpenAI-assisted application tailoring where configured
- 06Playwright auto-apply worker with rules-based Greenhouse auto-queue
Architecture
How the system fits together
02
MongoDB via Mongoose with User, Job, Match, and ApplyProfile models
03
Heuristic scoring service returning score + reasons for each job/user pair
04
OpenAI client for resume/cover tailoring when an API key is configured
05
Playwright workers for Greenhouse auto-apply against a rules-based queue
06
NextAuth for authenticated sessions; Zod validation on profile updates
Technology stack
Tools chosen for the problem
Next.js
TypeScript
MongoDB
Mongoose
OpenAI APIs
Playwright
NextAuth
Tailwind CSS
Screenshots
Selected frames from the product
Development process
How the build moved forward
- 01
Scoped an MVP around store → score → review — dashboard, jobs list, and profile first
- 02
Implemented scoring heuristics for role, skills, location, work mode, and seniority
- 03
Added apply profiles and best-profile selection for tailored applications
- 04
Wired OpenAI tailoring and Playwright auto-apply for supported Greenhouse URLs
- 05
Added ingestion/sync scripts, URL-quality cleanup, and digest/cron helpers
Lessons learned
What this build taught us
- 01Transparent match reasons build more trust than a score alone
- 02Auto-apply needs hard eligibility rules (provider, URL quality, score threshold) or it creates noise
- 03Multiple apply profiles beat one generic resume when candidates target different role families
Work shown is product and systems engineering — outcomes vary by scope and constraints.
Strategy
Want a build like Job Radar AI for your workflow?
Share the bottleneck in your business. We’ll map a comparable first phase — written scope, weekly demos, no invented promises.
More product builds
