All projects
Live · Multi-user 04 / Personal projects

ApplyBoard

Not a job board — the whole search, run for you. It finds the roles, scores the fit, writes the CV and the cover letter, and keeps the entire pipeline in one place. Cloud-native and built to serve many people, each in their own private instance.

"Searching for a job is a full-time job. ApplyBoard does the repetitive 90%, so the person only does the part that needs a human."
The problem

The job search is fragmented and repetitive: roles scattered across a dozen platforms, hundreds of irrelevant posts to filter, a CV and a cover letter to rewrite for every application, and no single place to know what you applied to, when, and what needs a follow-up. Most tools only solve one slice — usually just "find jobs". The other 80% of the work is left to the person.

The solution

ApplyBoard is the whole operating system for the search — not a feature of it. It covers the entire funnel end to end:

  • + Discovers roles across LinkedIn, Indeed, Glassdoor and company ATSs (Greenhouse, Ashby, Lever, Workday)
  • + Scores every role 0–9 against your profile with an LLM, and explains why
  • + Writes a tailored CV and cover letter for each role, automatically
  • + Centralises everything — applied, waiting, rejected, interviews — in one place
  • + Tracks follow-ups and surfaces the ones going cold
  • + A recruiter / hiring-manager contacts book for cold outreach
  • + One dashboard to run the whole search: filter, open, apply, mark, note
ApplyBoard dashboard: scored roles, a tailored CV and cover letter ready per role, apply-now queue and pipeline KPIs
The dashboard — anonymised, with sample data. Every role scored, CV and cover letter ready, one click to apply.
Built as a product

ApplyBoard isn't a script I run for myself — it's designed to serve many people. It's cloud-native and multi-tenant: every user gets their own private instance and their own isolated data, protected by row-level security in Postgres. Same engine, different person.

It already runs live for more than one profile, and the difference between them proves the point: one instance is tuned for AI & consulting roles, another for wealth management — different search criteria, different CV tracks, different scoring, different languages, fully independent. Onboarding a new person is spinning up their instance, not rebuilding the product.

What's built

A real, working system in daily use — thousands of roles processed and hundreds of applications managed through it.

Discovery engine
Pulls from the big boards plus direct ATS APIs of curated companies, so roles land the day they're posted — with the real application URL and the full description.
AI fit-scoring
Each role is scored 0–9 against the user's profile with reasoning, tuned per person with boost / penalty signals, seniority ceilings and language filters.
Document generation
Tailored CVs and cover letters per role, generated as PDF and DOCX from a set of base CV tracks and the user's source-of-truth profile.
The dashboard
A live app — local and online — with one-click actions: apply-now queue, open CV/letter, mark applied, star, skip, notes, response tracking and a follow-up queue.
Cloud backend
Supabase / Postgres with row-level security. Each user's data is isolated by user_id, synced between their machine and the cloud so the board is reachable anywhere.
Daily automation
A morning routine runs discovery, scoring and generation on a schedule and syncs the results — the board is fresh before the day starts.
Where it's going

The engine works and runs every day. The direction now is turning it from a multi-instance tool into a product other people can simply sign into: self-serve onboarding, a hosted dashboard per user, and moving beyond applications into the part that actually converts — warm outreach to the right people. The bet is that the winning job-search tool owns the whole funnel, not just the search box.