ve-work
Find the 5 roles that are real. Apply with intention.
A daily AI scout that finds the roles you'd actually take — on and off LinkedIn — researches each company and role, tailors your CV in your voice without inventing skills you don't have, and points you to the people inside who can warm-intro you.
Built on a simple read of the market: every senior role is becoming a forward-deployed role — and your search should run the same way.
The honest version: it doesn't apply for you. You still write the click.
Own it, don't rent it. Runs on your own VM (powered by OYASS) or locally on your machine — not someone else's cloud. The fork is a Claude Code skill; you own the stack. Or use the hosted instance at work.velab.org. Your LinkedIn data stays on your machine.
"The biggest 'aha' moment was not the discovery engine itself, but realizing that every morning I can simply open the queue and review opportunities. The value becomes clear once automation is running."
— early user, after 30 days
Every role is forward-deployed now
Palantir coined "forward deployed engineer": someone embedded in the customer's environment, building alongside them instead of shipping at them. The term is outgrowing engineering.
The senior roles worth taking now share one trait — the company isn't really hiring you to run the function. It's hiring you to deploy AI into the function. Forward-deployed accountant. Forward-deployed marketer. Forward-deployed ops lead. The job description won't use the words yet. The interview will be about nothing else.
ve-work reads roles through that lens. Not just title-matching — looking for the postings where the unwritten requirement is "be the edge of AI deployment on this team," because that's where senior hiring is converging and where your leverage is highest.
And there's a quiet proof built in: running ve-work is working forward-deployed. An agent does the legwork — on your own VM (powered by OYASS — Own Your AI Software Service) or running locally on your machine. You review evidence and make the calls. You walk into the interview already operating the way the role demands.
The honest version: almost no JD says "forward deployed." The thesis is a lens, not a filter — you'll still see normal-looking roles. The lens tells you which of them are secretly this.
30+ roles a day, on and off LinkedIn
Daily scan of LinkedIn plus company career pages and other job boards against your filters. AI triage drops the noise — recruiter reposts with hidden employers, wrong location, wrong seniority, your hard NOs. You see ~15 worth a look.
Know the company and the job first
A sourced profile of every shortlisted company — what they do, recent news, deals — plus a deep read of each job description: the real must-haves and where you match or have gaps. Every fact links its source, so you review with evidence, not guesswork.
CV + cover letter in 30 seconds
For any role you decide to apply to: a CV reordered around what that job actually asks for, and a letter that cites real company specifics — written in your voice. No invented skills, no template tone, no "I am writing to apply for".
Warm intros, not cold DMs
For each shortlisted company, real people fact-checked for a school/alumni tie and role relevance — ranked, each with a suggested intro angle. Not a "search LinkedIn for your school" link that leaves the work to you.
A living shortlist URL
Your password-protected page updates daily — roles, off-LinkedIn finds, company research, and your warm-intro contacts in one view. Different from a static PDF. Share with mentors or recruiters.
What it WON'T do
- Apply for you. Deliberate. Auto-apply is what gets you flagged by ATS systems and ghosted by recruiters. The "I read the JD and chose to apply" signal is the whole point.
- Fix a sparse LinkedIn. Better PDF in = better tailoring out. If your profile is two lines per role, the tailoring has nothing to work with.
- Find roles you didn't ask for. The keywords are the filter. Broaden them if you want broader.
- Cold-DM recruiters on LinkedIn. Against ToS, kills your account, doesn't work anyway.
- Promise interviews. It optimises for quality applications. Conversion is up to your background and the market.
Who it's for
You'll like this if
- You're a senior IC or operator (director+, head of, founder)
- You're searching with intent, not running a numbers game
- You value depth over volume
- You'd rather apply to 5 real fits than 100 generic ones
- You trust me with your LinkedIn data
- You read "deploy AI into the team" as your real job description, whatever your title says
- You want to own the stack — run it on your own VM (powered by OYASS) or locally, not rent someone else's cloud
Skip this if
- You want to send 200 applications this month
- You're looking for entry-level or new-grad roles
- You want a recruiter, not a tool
- You expect interviews-as-a-service
- You think AI is a feature of your tools, not the substance of your next role
The morning ritual
This is what success looks like from day one — not the discovery engine, not the tuning knobs.
Every morning you open your queue and ~10–15 roles are waiting — researched, filtered, ready. That's it. You don't tune watchlists, ATS coverage, or scoring weights on day one. An early user told us they spent too long on optimization knobs before seeing any output — so we changed the design: get to an automated baseline first, optimize on outcomes later. The knobs are there when you have data to turn them against. On day one, there's just the queue.
What happens every day automatically
- Discovers ~30 new LinkedIn job postings matching your criteria
- AI-filters out the obvious junk — commission-only, internships, wrong seniority, wrong country, your hard-NO terms
- Validates each listing is still accepting applications (no dead links)
- Surfaces a one-click "Find <your school> alums at this company" button per row — opens LinkedIn search in your own logged-in session
- Tags any role you've already tailored with a match score
- Updates your private dashboard + shareable shortlist
You wake up to a shortlist of ~10–15 roles worth looking at. Junk doesn't reach you.
How to actually apply
- See a role on your dashboard or in the morning ping
- Decide you want it
- Tell me the URL (or ask Claude) → CV + cover letter tailored in ~30 seconds — actually reading the JD, mapping your real experience, not pattern-matching
- You paste/upload to LinkedIn yourself
The CV gets reordered to lead with what's most relevant to that specific role, written in your voice, never inventing skills you don't have. The cover letter is 3 paragraphs: why this company specifically, your most relevant proof, a specific ask. No "I am writing to apply for" or "I believe I would be a great fit" — those are banned.
Install the Claude Code skill
Download & extract
mkdir -p ~/.claude/skills
tar xzf ~/Downloads/ve-work-self-serve.tar.gz -C ~/.claude/skills
chmod +x ~/.claude/skills/ve-work-self-serve/vework.shRestart Claude Code so it picks up the new skill.
Log in
Use the credentials from your invite email:
bash ~/.claude/skills/ve-work-self-serve/vework.sh login "<your-email>" "<your-password>"Build your profile (first run)
Save your LinkedIn profile as PDF (on linkedin.com: More → Save to PDF), then:
bash ~/.claude/skills/ve-work-self-serve/vework.sh onboard ~/Downloads/Profile.pdf \
keywords="product manager, solutions consultant" \
locations="New York, Remote"Takes a few minutes. Then just ask Claude things like "check my job matches", "refresh my search", "star this role", "I applied to this".
Self-hosting? Set VE_WORK_BASE to point the skill at your own deployment:
export VE_WORK_BASE=https://your-vm.exe.xyz # OYASS deployment
# or
export VE_WORK_BASE=http://localhost:3000 # local runHosting with me? The default (work.velab.org) just works — no env var needed. Hosting yourself? You own everything: the fork, the pipeline, your data. OYASS gives you the VM template to deploy in minutes.
What the skill can do
| Intent | Command |
|---|---|
| Show new matches / shortlist | matches |
| Pipeline status / last run | status |
| Cheap re-rank (no new discovery) | refresh |
| Full fresh search | discover |
| Star a role | star <url> |
| Dismiss a role | skip <url> |
| Mark applied | applied <url> "Title" "Company" |
| Add a job found elsewhere | external <url> "Title" "Company" |
| Update preferences | prefs keywords="..." locations="..." |
All the same actions as the web dashboard, server-gated to your account only.
Privacy & ownership
- Runs on your own VM (powered by OYASS) or locally — not on someone else's cloud
- The fork is a Claude Code skill — you own the stack: run it, modify it, or walk away with everything
- Your LinkedIn PDF stays on disk, never committed to git
- Your dashboard is password-protected (your email + a password I send you)
- Your shareable URL is password-protected (separate, easier-to-remember password)
- Self-hosted = I see nothing. Hosted with me = I can read your profile to help you debug. Nothing leaves unless you share your URL.
- No analytics, no telemetry, no third-party trackers.
Cost
- Free for you. If you host with me, I cover infrastructure (Apify discovery + LLM tailoring + hosting). If you self-host (your own VM via OYASS, or locally), you cover your own Apify + LLM costs — the fork and skill are free.
- Your time: ~5 minutes once to onboard, ~30 seconds per role to decide if you want it tailored, ~5 minutes per real application to paste into LinkedIn and click submit.
To start
Text or email me your LinkedIn PDF + 3–8 keywords you'd take. I'll set you up tonight. First results show up on your dashboard within 24 hours.
This is a tool for being deliberate, not for scaling generic applications. If your goal is "send 200 applications this month", this is the wrong tool. If your goal is "find the 5 roles that are real and apply with intention", this is exactly the tool.
— Alet