One profile powers your commercial, impact film, live opportunities, and a bias-aware applicant ↔ company career-fair match built around skills, role requirements, culture and mutual fit. Recruiter information is optional.
ChatGPT merges your source material
ElevenLabs Scribe transcribes your spoken story
ElevenLabs Voice performs the commercial
Eleven Music creates the background track
Type it, say it, or both — like you'd tell a friend: where you've been, what you can do, where you're headed. Add your resume if you have one. No resume? Your story is enough.
The original file stays in your browser. When you run an AI feature, extracted text may be sent to the connected AI service so it can build or evaluate your profile.
Provider / staff means your company uses AppYOU to serve paying or sponsored clients. Self-service means the person or business uses the same engine for themselves. This is separate from the profile type below.
Paste the HTTPS endpoint where your Muse agent receives work (n8n, Make, Zapier, your server or agent API). Each approved application is POSTed as one JSON packet: resume or capability statement PDF (base64), cover letter, opportunity facts and the approval record. Muse replies {status, confirmationId, message, questions, statusUrl}. The endpoint must allow requests from this app (CORS). No URL? Approved packets download for hand-off instead.
For real participants, route OpenAI and ElevenLabs requests through your organization's server so vendor API keys do not live in the browser. The gateway should authenticate the signed-in user, add vendor keys from a secret manager, enforce server-side budgets, and write an audit log.
/v1/models and adds general GPT/o-series models your key can access. “Test selected model + web search” performs one small live web-search request and therefore uses API credits. The selected override is used across resume mapping, research, opportunity discovery/verification, company research, Mutual Fit, Learn 2 Earn, SROI narrative, Career Journey and commercial writing.Tracks actual provider usage by session, participant and month. Enter your current contracted rates; the engine intentionally does not hard-code vendor prices. Session, participant and monthly limits can stop a paid OpenAI request before it leaves the browser.
Pre-embedded demo keys (if set in this file's PREEMBEDDED_KEYS block) connect automatically on load. This file is served from a public URL, so only monitored/rotatable demo keys should ever be embedded — never a production key. Keys you paste are kept for this tab (they survive a refresh) unless you tick Remember keys on this device. Clear keys removes them from this tab and this browser.
PDF, Word, PowerPoint, Excel, CSV, RTF or plain text. The original file stays in your browser. Extracted text is added to your profile and may be sent to connected AI services only when you run an AI-powered feature.
Paste your public LinkedIn profile URL, or upload a screenshot/image of your LinkedIn QR code. GPT will use public web-visible professional information to prefill this profile; you review and edit it before anything is used in your commercial.
Only public professional information is imported. LinkedIn visibility settings can limit what is available; if the profile is not publicly readable, upload your resume or LinkedIn PDF instead.
First the engine audits your uploaded resume the way employer AI screeners read it — parseability, keywords, missing metadata. Then it builds a tailored, ATS-safe resume for the jobs it found, from your resume and your story — typed or voice. No resume? Your story alone is enough to build from. Grounded only in your real experience; nothing is ever invented. The tailored resume always builds in English for US employer/ATS screening. The Google Translate control changes the page interface only.
For RFP / RFI / grant responses, the checker also uses the loaded solicitation and compliance matrix when available.
The PDF prints exactly what is in the box. Keep the layout: ALL-CAPS lines become section headings, Employer — Title | Dates lines get the date right-aligned, lines starting with “•” become bullets, and “Heading: detail” lines get a bold lead.
Optional, and entirely yours to control. Nothing here is guessed from your address — only what you tick is ever used, and you decide how much of it shows up.
Framed as capability, never as pity: the obstacle gets named once, and the rest of the words go to what you can now do because of it.
A 15–60 second spoken pitch in your voice. Do this after your resume — it uses the same profile.
Record a sample here, or reuse the story clip from the left. For the most consistent clone, aim for 60 to 120 seconds of clear speech with no music behind it.
The selected voice sample is sent to ElevenLabs to create the clone. Unless you delete it here or choose automatic deletion, the cloned voice can remain in the connected ElevenLabs account.
Short tracks loop automatically to fill the spot. Make sure you own or are licensed to use the audio.
Recorded through the mic with noise suppression and auto-gain switched off, so a live instrument or a track playing in the room comes through clean. Get the mic close to the source and keep the room quiet.
The spot ends exactly this long after the last spoken word — no leftover music. The script target above shrinks to match, so the read fills the rest.
Every word here gets spoken, exactly as written. Let ChatGPT draft it, then edit freely — or skip ChatGPT and write it yourself.
The written case behind the commercial — positioning, proof, objections, gaps to close, prescribed outcomes, a sales-resume summary and the ask. Built for one reader at a time.
Your photos, clips and audio testimony are organized into an AI-assisted executive edit with captions, titles, optional logo treatment, source-audio ducking and a timed final soundtrack. GPT can inspect compressed photos and sampled video frames to choose stronger shots.
For Executive Produce, the app sends compressed still images and sampled video frames to GPT so it can understand the actual visual content. Raw full-resolution media stays in the browser.
Most people watch with the sound off. Burned-in captions are the difference between a film that lands and one that gets scrolled past.
The impact story is the funding case. Ask the participant for their pictures, videos and needs list, shape their story, price what it takes — then print the one-pager a funder can say yes to.
Send this ask by text or WhatsApp, then drop what comes back into Shots above and the story fields below.
Total ask: $0
Press one button. Every job is checked live before it is shown, ranked by how well it fits you, and newest postings come first.
Leave blank and it uses the hometown or organisation from your profile. The search names the towns inside that radius and goes through the largest employers in the area one by one, not just the job boards.
For jobs, internships and co-ops, the engine first maps real employers with a worksite inside your radius that hire for your target roles, title/major, skills and story, then opens each one's own careers page or applicant-tracking board (Workday, Greenhouse, iCIMS…) and pulls the matching open roles. The coverage list shows every employer mapped and whether its careers page was reached. Run "Find more" to sweep the next set of employers.
Tick what matters to you. Each match is then checked against what the employer publishes and what current and former staff say publicly, and you get told which of your needs are met, which conflict, and which nobody could confirm.
Employer reviews are opinions, not facts, and policies differ by role, site and state — marijuana screening is still legally required for some safety-sensitive and federally regulated jobs. Each result says what was actually found and what nobody could confirm, so you can ask in the interview rather than assume.
Measure the journey from the participant's real resume today → verified opportunities they can pursue now → prescribed official Learn 2 Earn topics/modules and, when relevant, the Cut2DaCASH financial-literacy extension → a projected post-training profile → a second verified opportunity search → actual completion and employment/business outcomes over time.
GPT compares the baseline profile to requirements and gaps in the verified opportunity set, then maps those gaps to the official 7-module / 35-topic Learn 2 Earn curriculum. Module 8, Cut2DaCASH Financial Literacy, is an extension pathway and is prescribed only when the evidence shows a financial-capability, wealth-building or business-readiness need. Non-course barriers stay separate.
The official Learn 2 Earn core is 7 modules / 15 required hours with 5 topics per module. Module 8 is the Cut2DaCASH Financial Literacy extension and is not counted inside the 15 required Learn 2 Earn hours. Topic-level curriculum is fixed to the official catalog; module completion, pre/post assessment and evidence remain tracked separately.
The second search uses the exact same verified research engine and geography, but with the post-training scenario layered onto the baseline. New opportunities and fit changes are therefore visible without overwriting the person's current resume.
Rebuilds the participant's resume from everything they had before plus what the program produced: the Learn 2 Earn / Cut2DaCASH certification, the skills proven by module assessments, and any job or contract they started (from Career Journey). Uses the same resume builder and PDF styles as the Profile tab.
What the Learn 2 Earn training produced for this participant: the job or contract obtained (from Career Journey), the extra wages or revenue it brings compared with before training, and the share of that increase the training earned. This drives the training ROI and SROI.
Builds benefit factors from the prescribed modules and the before/after opportunity results: better-paying high-fit jobs, and vendor/supplier contracts or grants for small businesses. It also adds program evidence and suggested follow-up ratings to each SDOH domain. All of this is projected and kept separate from verified outcomes.
Pulls cited figures for this participant's place and target roles — Census ACS, BLS wages, Opportunity Atlas (Chetty), CDC PLACES, FTC fraud data, CFPB/FINRA, Census Nonemployer Statistics, city/county impact zones and corridors, school-district data and the city procurement portal — and uses them instead of the defaults. Each figure keeps its link, year and geography.
Placement likelihood, win rate and margin are planning assumptions — set them from the participant's own bid history or the program's cohort data when you have it.
Wage ROI uses the incremental annual wage gain (follow-up minus baseline), not the full salary. It enters verified ROI/SROI only when status is Verified actual and evidence is supplied.
Leave attribution or deadweight blank until the evaluator/funder chooses defensible assumptions. Gross verified value still calculates; adjusted SROI waits for the assumptions.
Mapping self-declared context only adds evaluation notes; it does not guess a baseline score. The participant/evaluator still chooses the 1–5 baseline and follow-up rating.
Rate only assessed domains: 1 = severe barrier, 2 = material barrier, 3 = mixed/unstable, 4 = mostly stable, 5 = stable/strong access. Standard SDOH domains are supplemented with Digital Inclusion and Financial Capability because those are direct program pathways.
Imported Career Journey amounts are excluded from the calculation by default so a salary, grant ceiling or contract value cannot be double-counted. Review the amount, classify it as incremental benefit, then check include.
Close the loop from opportunity → application / submission → interview / review → offer / award → start / contract → 30/60/90-day success. This layer reuses the same verified opportunity, applicant profile, company role and Learn 2 Earn evidence already in the engine.
GPT turns the active opportunity into an evidence-based action plan. Missing resume wording is separated from missing capability; Learn 2 Earn modules may be referenced only when they actually address the gap.
Creates applicant preparation and a standardized employer guide from the same role evidence. The employer guide supports human judgment; it does not rank, reject or make a hiring decision.
Completed Learn 2 Earn / Cut2DaCASH modules synchronize here automatically. A completion without assessment/evidence remains marked pending proof and is not treated as a verified skill claim in GPT career guidance.
A score by itself is not verification. Only Provider verified credentials with supporting evidence are treated as verified in GPT guidance and KPI reporting.
Once the active opportunity reaches an accepted, started or contracted stage, GPT creates a role-specific onboarding and success plan: expectations, skills, manager checkpoints, mentor/support needs, training milestones, performance evidence and employee self-check-ins.
Company / recruiter view: build the Company Passport from the employer website, load current openings from official careers/ATS pages, then compare the applicant's existing profile to those real roles. Recruiter profiles remain optional; the role and company profile are the primary match inputs.
Enter the website first. GPT can research the official company site, populate the Company Passport, find the careers page and load current roles automatically.
Prototype storage is local to this browser. Export the Company Passport JSON to move it to another device. A production version should use authenticated, role-based storage.
Optional. Add recruiters only when the company wants named routing at a career fair. Their professional background can route applicants to the right person; personality or working-style information only helps shape the conversation and never changes qualification or fit.
The company website/ATS is the source of truth. GPT first discovers the company's live openings without using the applicant profile to bias what exists. The engine then re-verifies each posting, filters and sorts the verified roles against the same applicant metadata/content used by Applicant Opportunities, and sends only plausible roles into Mutual Fit.
A role is never invented to improve fit. Missing role data stays unknown; explicit conflicts can exclude a role. If no strong roles exist at this company, the engine says so instead of forcing low-quality matches.
Manual roles are preserved when GPT refreshes the careers site.
These fields are the structured applicant record. A resume upload can populate them automatically, and the Mutual Fit engine reads the same values.
Entirely optional, and it does not touch your match. It is never sent to the matching model, never shown to a recruiter, and is not stored inside the participant profile or Career Journey. It is attached only to the separate fairness-audit record after scoring and clears when you switch participants or reload the page.
Deciding what good looks like after meeting candidates is where bias enters. Set the criteria first and lock them; every applicant is then read against the same list, and any later change is recorded with its reason.
For the career centre, after the event. Collect the assessment records from every table, load them here, and see whether any group advanced at a materially lower rate. One record per applicant, aggregated on this device.
Manage active clients, service delivery, job / opportunity progress, follow-ups and outcomes without opening every record one by one.
| Client | Persona | Service | Payment | Pipeline | Readiness | Opps | Journey | Assigned staff | Follow-up | Next action | Last activity |
|---|
Workspace backup: downloads client/workflow records from this browser. API keys, secure-gateway credentials and tokens are intentionally excluded. The file contains client data — store it securely.
Use the same engine yourself. Your profile, resume, opportunities, matches, applications and outcomes stay connected.
One place for the active client's or self-service user's finished files.
Owner/admin view. Inspect and edit the exact prompt templates used by the research and matching engine. Prompt changes are saved in this browser, versioned on export, and applied on the next run. Code-level source verification, identity redaction and structured-output gates remain active even when a prompt is edited.