Alumni

Serper 0 · Haiku $0.000
APIs not configured

Alumni Intelligence

Import, verify, then capture exact LinkedIn Experience through the local helper.

Recent imports

System status

Imports

Upload a batch list and turn it into verified alumni profiles.

Re-enrich all Destructive

The Enrich YES Profiles button above never re-reads someone already enriched. This does the opposite: it re-scrapes people you already have and overwrites their captured employment history with whatever LinkedIn serves today. Use it when an extraction bug has been fixed and the records written by the old extractor need re-reading. The previous history is archived first and can be restored, but the visible record changes immediately.

The token is the one stored as ADMIN_TOKEN in Cloudflare. It is held in memory for this tab only — never saved to the browser — so you will be asked for it again after a refresh.

Human Review

Resolve ambiguous identity matches and low-confidence domain classifications.

LinkedIn matching

Domain classification

Alumni Directory

Search and filter the shared alumni database by latest role.

Shared database

Favourites

Every alumnus you starred, with the full record and your comment shown inline.

Shared database

Helper Monitor

What the enrichment helper is doing right now, straight from the job queue.

Live
Loading…

Settings

Search configuration, AI enrichment and shared database controls.

Serper API

Serper-only web research: all external career data comes from Serper/Google results. No external profile scraper or browser automation is used. Haiku only structures the Serper evidence and classifies ambiguous job functions; it does not search the web.
Saved in this browser until you update or remove it. The key is sent only to the Cloudflare Serper proxy when a search runs. Large Serper result payloads are no longer stored in browser storage, preventing quota failures on big batches.

Anthropic Haiku API

Server managed: the Anthropic key is stored securely in Cloudflare and is not exposed in this browser. Haiku is used only for ambiguous identity judgement and compact company/role-history structuring.

Accuracy-first Serper search

Low-credit role-history enrichment: after you confirm Yes, the app reuses the existing discovery evidence and uses a deep company/role search: 4 broad searches, career-window searches across the person's working years, then 1 targeted search for each employer found. It intentionally skips dates and descriptions so the extra searches are spent only on recovering complete company + role history.
Default 7. Higher uses more free Serper credits, not Claude web-search spend.
Deep role-history mode. Search count varies by career length and number of employers. All requests stay at 10 results and avoid restricted operators. The app automatically simplifies any query rejected by Serper's free tier.
Haiku compares candidates only when top deterministic scores are close or evidence conflicts.

Cheap Haiku controls

Profile discovery uses Serper. Enrichment copies the complete authenticated LinkedIn Experience section locally, then one inexpensive Haiku call structures every company, role/promotion, date, location and description.

Usage & cache

The Serper counter is local and starts from 2,500 free credits by default. Check the Serper dashboard for the authoritative remaining balance.

Shared database & backup

Shared storage: alumni data is saved to Supabase through Vercel. A local browser copy is kept only as a fallback/cache. JSON backup and restore still work.

Activity log