The boring answer undersells it a little, so here’s the long version.

How it started

October 1st, one commit: init: career vault for Spring–Summer 2027 internship campaign.

The idea wasn’t new for me. I’d been running my actual code repositories this way for a while, with notes the agent reads before it touches anything. This is the same habit, pointed at my own life.

The first real step was a dump. I basically had ChatGPT and Claude export everything I’d done with them over the past three or four years, put that next to every resume I’ve ever written, and turned it into one profile. So I stopped rewriting who I am for every application.

Twelve days later

The zero is on purpose. It’s a one-person repo and the agent is the only other contributor, so the hooks commit and push as I go. My job hunt has a commit history instead of a review queue.

How it works

• One note per thing. One per company, one per application, one per person. Postings and emails get saved exactly as they arrived and are never edited.
• The agent does the filing. When something new lands, it links the posting to the company, the company to the people I know there, and updates the pipeline. I make the decisions.
• Every resume is curated for every application. The agent reads everything a posting asks for, then looks at everything in the profile and builds the best version of me that fits the role. I’d rather apply to roles that are exciting for me than mass send.
• Nothing goes missing. Any conversation I have about the search gets logged into the repo, so there’s nothing I could say that wouldn’t be in there. Ask it anything and it gives you a full status update.
• GitHub, because it’s so easy. It’s the de facto way of doing this, it’s already wired into everything, and markdown is just text.

What plugs into it

• Claude Code runs it, with 8 slash commands. The four I use every day are /apply, /company, /interview, and /pipeline.
• Git hooks stage every edit, then commit and push when a session ends.
• Gmail for recruiter emails and application receipts.
• Granola for call notes.
• Chrome for reading postings and keeping LinkedIn up to date.
• PostHog for what people actually do on this site.
• Vercel, Supabase, and Resend behind the site, the helper portal, and the emails.

What grew out of it

The folder was supposed to be the whole project. Then it kept spawning things:

• This site (October 6). A very fun responsive way to get to know my life story professionally, from the Rice lab to now. The “How I think” section is the folder itself.
• A kit (October 10). People asked how to set this up, so there’s an empty copy you can run: npx career-vault, or build one from the kit on my homepage. Nothing of mine comes along.
• A helper portal (October 11). A small private page where family and friends each get their own link to the roles I’m going after and their req numbers. If a name comes to mind, they can see what’s relevant in thirty seconds instead of texting me back and forth. It reads straight from the folder, and anything they claim flows back into it.

Why bother

For the roles I want, they want to see how your brain works. This is the most honest version of that I could think of: the real system, running, with the commit history to prove it.

Also it’s fun. I get kind of giddy every time the agent files something correctly.

I’m looking for Summer 2027 internships in product, TPM, and AI solutions. If something here makes you think of a team, I’d love to hear about it at contact@emrecanselcuk.com.