Frequently asked questions

ve-work is VE Lab’s job dashboard. It discovers roles and ranks them by fit to your profile. It also builds company dossiers and surfaces warm intros (alumni / ex-colleagues). An AI assistant answers questions and acts on your shortlist. It runs on work.velab.org.

It opens the ve-work Assistant. The chat works on your own data: the shortlist, fit scores, company dossiers, and warm paths. Ask things like “what are my top 5 matches?”, “which roles have warm intros?”, or “what do you know about [company]?”. It can also act for you: star, skip, mark a role applied, or kick off a pipeline run.

Only within limits. It can read anything in your profile folder. It can also take the same queue actions you can — star, skip, mark applied — and trigger a refresh. It cannot delete or rename files. It cannot edit your search.yaml or profile, or write outside your own folder. Those stay manual on purpose. Queue changes go through the same API the buttons use, so nothing happens off the record.

The assistant only works on your own profile. You must sign in, and you can only chat about your own data. The agent process runs on the ve-work server, scoped to your profile folder. Ollama Cloud generates the model responses. Your questions and the profile context the assistant reads go there for processing. Treat it like any other cloud AI tool: avoid pasting anything you would not send to a third party.

We keep a working shortlist so the dashboard stays usable. Your max_results in search.yaml sets the cap — 30 on a stock profile, 40 if unset. Each run also keeps max 2 roles per company, and fit ≥ 7 when a score exists. Extra roles move to History / Processed — not discarded forever.

That is a second filter, separate from the cap above. The Shortlist tab hides roles scored 3 or below, plus anything you skipped, to keep the list signal-heavy. It never hides starred or applied roles. The count tells you they exist — the roles are still in your data.

Several sources: LinkedIn, company ATS boards (Greenhouse, Lever, Workday and similar), web search, and RSS feeds. Anything you add by hand from a careers page lands under the Other sources tab. That tab stays empty until you use it.

Two reasons. First, each run scouts only the top ~12 shortlist companies. A company can therefore appear under Companies before we search it for intros. Second, a path only counts for a confirmed match to your schools or past employers. The person also needs a real name and LinkedIn profile. We filter out wrong-school noise on purpose. Companies is also capped (~40) and skips staffing and recruiting firms.

The chips are All time · Last 48h · This week · Last week. They parse the relative “posted X ago” text scraped from the job listing (e.g. LinkedIn). Roles without a readable post date only show under All time.

The automated pipeline typically runs around 07:00 and 19:00 UTC. You can also trigger discovery or enrichment from Pipeline in the account menu (⋯).

Yes. Account menu → Shareable shortlist gives a read-only URL under /work/s/… with the same roles, fit, and warm paths. The password is optional — and off by default. Set publish.password in your search.yaml to protect it. Until you do, anyone with the link can open it — treat an unprotected URL as public. The share panel tells you which state you are in.

Use Edit profile in the account menu (⋯) — it edits your search.yaml directly, which is where keywords and locations live. (Preferences asks for those only during first-run onboarding. An existing account will not show the fields there.) You can also ask an operator to change them. After a search change, prior shortlists can appear under History.

ve-work is part of VE Lab. Product and ops questions: reach your VE Lab contact or start at velab.org.