Why cvme Runs Entirely in Your Browser — No Servers, No Uploads
Your resume is one of the most personal documents you own
A single resume typically contains your full name, phone number, home city, employment history going back a decade or more, education, and often links to personal profiles. It's exactly the kind of document you don't want sitting on an unknown company's server indefinitely, especially when the only reason it's there is that you wanted help formatting it. That mismatch — a low-stakes formatting task requiring you to hand over a high-sensitivity document — is the problem cvme is built to avoid.
What "runs entirely in your browser" actually means
Most AI-powered tools work by sending your input to a server, running a large model there, and sending a response back. cvme does something different: it downloads a compact AI model directly to your device the first time you use it, and every extraction after that runs using your own device's hardware, through your browser. Your resume's text never leaves your machine to be processed. There's no upload step, no server-side storage, and nothing to delete later, because nothing was sent anywhere in the first place.
Why this makes it offline-capable too
Once the model has been downloaded once, it stays cached in your browser. That means cvme keeps working even without an internet connection — useful if you're polishing a resume on a flight, in a coffee shop with unreliable Wi-Fi, or anywhere else connectivity isn't guaranteed. A tool that depends on a round trip to a server can't offer that; one that runs locally can.
Choosing a model size
cvme offers a small set of local models to choose from, all built on the same Qwen2.5 architecture at different sizes — roughly 0.5B, 1.5B, and 3B parameters. The trade-off is straightforward: the smallest model downloads fastest and runs comfortably on almost any device, while the larger models take longer to download and need more memory but tend to produce more accurate extraction on complex or unusually formatted resumes. There's no wrong choice — it's a dial between speed and quality that you can adjust based on your device and how much of a rush you're in, and the picker remembers whatever you last chose.
The trade-off, honestly
Running a model in-browser does mean the first use involves a real download — typically under a gigabyte for the lightest option — and processing happens on your device's own hardware rather than a data center's. On modest hardware that can mean extraction takes a bit longer than a cloud round trip would. In exchange, your resume's content stays under your control the entire time, and the tool keeps working with no connection at all once that first download is done. For a document as personal as a resume, that trade feels like the right one to make by default.
What this doesn't change
Local processing doesn't make the rest of cvme any less capable — you still get full field-by-field editing, multiple templates, a chat-based editing bar, and one-click PDF export. The only difference is where the AI work happens: on your device, under your control, rather than somewhere else.