JobNet Application · Career pipeline · Preview
A pipeline that thinks
before it drafts.
One app for the whole loop. JobNet scrapes public LinkedIn and Glassdoor listings, ranks them against real CV evidence, and AI Hawk (formerly Drop Net) drafts a tailored cover note and CV synopsis for the role you pick. You review it, then copy it or forward it by email yourself. In the real pipeline drafts land in Gmail as drafts. Nothing is ever sent or applied for automatically.
Freelance mode points the same models at small businesses: it scores which ones have a real AI-automation problem worth solving, suggests the offer, and drafts outreach that a person reviews before anything is sent.
- 1 · ShortlistLIVE
- 2 · Matchon-device
- 3 · AI Hawk draftDEMO · templates
- 4 · Reviewyou
- 5 · Handoffcopy · forward
- 8live ML features
- 0third-party calls
- 0auto-sends, ever
- 100%runs on this device
Shortlist & explainable match
The shortlist comes from JobNet's read-only public API (the scheduled scrape runs off this site). Sentence embeddings (MiniLM, self-hosted, in a background worker) compare each requirement with real CV evidence; BM25 takes over instantly when the model isn't loaded. Pick a role, then send it straight to AI Hawk.
Paste any job description
AI Hawk draft
The draft step that used to be the separate Drop Net page. AI Hawk writes a cover note and CV synopsis for one role, built only from facts on the published CV and case studies, with sources. Anything it can't back up stays a [placeholder] or is listed as a gap. It runs from templates on this device: no LLM call, nothing stored or sent.
CV order for this role
In Katlego's own pipeline this step becomes a Gmail draft for him to check and send. This public page only copies the text or opens your own mail app.
Toolkit for the selected role
Fifteen tools around the draft: an honest ATS readability check, salary hint, interview prep, follow-up planner, duplicate finder, timeline, tone, PDF, share link, offline cache, GitHub stack and repo look-ups, and a per-role radar. Press Ctrl/⌘+K for the command palette or Ctrl/⌘+J for JARVIS mode.
Loading the toolkit…
Smart prioritiser
Priority = fita × freshnessb × salary signalc. Freshness halves every 5 days; undisclosed pay gets a neutral 0.7. Drag the weights.
Skill-gap radar
Market demand across the current listings vs. evidenced strength (shipped = 1, partial = 0.6, listed on CV = 0.45, not evidenced = 0). The biggest gaps become a learning plan with a proof-of-work project.
Learning plan
Application-outcome model
A logistic regression that estimates the chance of a positive reply from how and where an application went. Toggle the inputs to see each feature's contribution.
Inbox reply classifier
Multinomial Naive Bayes (words + bigrams), trained in your browser on a small labelled set, then blended with embedding similarity when the model is loaded. It labels each reply as interview, rejection or acknowledgement and shows the cues it used.
Sample emails written for this demo. No real inbox content, senders or recruiter names appear on this page.
Freelance-lead mode
Scores small businesses for AI-automation fit: repetitive pain, matching service, budget, size and digital readiness. It suggests the offer and drafts outreach for a person to review. It never sends anything.
Describe a business
Pipeline dashboard
The daily loop at a glance: funnel, drafts waiting on a human, and the nudge clock. The numbers below are sample data that mirrors the real flow.