Turn a conventional CV into evidence for a Canadian Tri-Agency narrative CV.
The assistant helps you find your role, relevance, quality, impact, and
mentorship evidence before it drafts. You remain the author and fact-checker.
Why local AI?
Your CV is processed by a small model running on your device. Its extracted text and drafts are not sent to an AI provider or this site's server.
The trade-off
Local models are slower and less capable than leading cloud models, require a large first download, and may miss context. Your CV content and drafts stay on your device and are not sent to a cloud model.
What the tool adds
Agency-informed prompts, application-specific selection, evidence-gap checks, and anti-invention rules help you find the right material—not merely rewrite a publication list.
WebGPU is required for local processing.
Download the local AI model
Start with the model selected for your device. You can add your CV and
application details while it prepares. Keep this tab open; your CV is
processed on your computer.
Checking whether this model is already saved in your browser…
The first download is approximately 0.9 GB for the selected model.
Use Wi-Fi and keep this tab open while it prepares.
Preparing the selected model…
Measuring preparation time…
Saved files can skip the download. The model still needs time to get ready on your device.
Model diagnostics
Source upload and agency questions are available while the model loads. AI suggestions, extraction, and drafting become available when it is ready.
Model storage and advanced help
The first use downloads model files into browser storage. Later visits can usually reuse them. Clearing those files requires a new download.
The models are quantized to reduce their size. Larger options need more storage and graphics memory. Additional compatible models are available in Settings.
Choose the funding agency
Select one planning lens. The next screen will use it to ask agency-specific evidence questions.
Agency-specific evidence
Review the prompts for your selected agency. The local model can suggest answers from your source, but you remain responsible for verifying every claim.
How would you like to begin?
All four routes lead to the same evidence review. Choose the source that gives you the best factual starting point.
Select your CV
Choose a text-based PDF, Word (.docx), or plain-text file.
Scanned image PDFs are not supported in this version.
Saved document text stays in this browser only.
Optional: rank the CV elements that matter most for this application. The original CV stays unchanged; extraction uses a shorter priority view to reduce token-window pressure.
Fast mode is recommended for smaller local models and long CVs. You can complete optional context during review.
Preview the priority context used for analysisPreview extracted textQuestions and suggested wording for application context
What is the funding opportunity and which evaluation criteria matter for your role?
What problem, knowledge gap, or community need will the proposed work address?
Who might use or benefit from the work, and through which planned activities?
What will you personally lead, contribute, or be accountable for?
Try: “For the [opportunity], the project will address [problem] with/for [partners or users]. In my role as [role], I will be responsible for [specific responsibilities]. My prior [expertise or experience] is especially relevant to [criterion or project need].”
Verify extracted evidence
Review what the model found, then complete the evidence pathway from your
contribution to its use and outcomes. Do not leave invented facts in place.
Preparing extraction…
0 sec elapsed
If interrupted, resume the remaining analysis below. Completed results and your corrections stay available in this page; turn on browser saving to keep them after reloading.
Generate a thematic section
Choose your discipline so importance ranking and drafting use the right
evidence signals. Confirm the top elements, then generate one section.
Step 1 of 4 · Choose the discipline and TCV section.
Most important elements
Ranked instantly by discipline signals. Local AI refinement is optional.
0 of 10 selected
Preparing draft…
0 sec elapsed
Partial text is retained if generation is interrupted. You can edit it or generate again.
