TalentoTalento
Built for targeted applications.
The ATS analysis you will never do by hand again.
Every application starts from scratch: re-reading the job offer, guessing the keywords, rewriting the same sentences. Talento locks in the decisions — extraction, semantic matching, scoring, interview prep — and replays them on every offer in a single pass.
For people aiming at one specific job, not fifty.
reading offer.pdf … ok
reading resume.pdf … ok
extracting criteria … 18 found
generating rewrites
Common problems
Tailoring a resume to a job offer costs a day. Every time.
It is never the easy part that hurts. It is re-reading the offer line by line to guess what the filter expects. It is rewriting the same experience under a different job title. It is the PDF the ATS parses badly, and the automatic rejection before a human ever opens the file.
Nobody counts those hours, and everyone pays them again on the next application.
Architecture
Every decision already made. So you can go straight to the interview.
A resume builder gives you templates. Here, you get decisions: how the offer is broken down, how a skill is recognized as equivalent, which gaps matter, in what order results reach the screen. Set once, applied to every analysis. That is what makes the model useful: an unconstrained LLM improvises, an LLM inside a fixed structure produces results you can compare from one offer to the next.
A single pass
The offer and the resume go out together in one structured call. No prompt chain, no stacked latency: the first answer arrives in under two seconds.
Semantic matching
“Vue 3” and “Vue.js”, “Postgres” and “PostgreSQL”, “fluent English” and “bilingual” are recognized as equivalent. Intent matters, not the exact string.
Five gap categories
Technical skills, soft skills, languages, experience level, degrees and certifications. The same boxes on every analysis, with the same matching threshold.
Ready-to-paste rewrites
Every weak or missing criterion gets a results-oriented wording, readable by an ATS and a recruiter alike, that you paste straight into your document.
Interview built from the gap
Questions do not come from a generic bank: they are generated from the exact difference between your profile and the offer, with feedback on every answer.
Results as they stream
Criteria appear one by one through an SSE stream. You read the first results while the next ones arrive, with no loading screen.
Data control
Files are not archived beyond the session. Gemini receives the text for the analysis; signed-in accounts can keep a history. No resume is ever sold, and you can delete everything from your profile.
Real preview
This is the tool, not a mockup.
Upload
Your resume as a PDF and the job offer text. The text is extracted and cleaned up automatically.
Read
Overall score, per-criterion details, ranked gaps and rewrites, displayed as they stream.
Prepare
Ten interview questions built on your actual weak spots, with instant feedback.
Access
Free to try. Open source.
The project is released under the MIT License. You can use it online, read the code, host it yourself or contribute.
- Trial without a credit card
- Full analysis, streamed live
- Ranked gaps and rewrites
- Interview simulator with feedback
- PDF import and automatic extraction
- Open source, self-hostable
Before you start
Questions people ask us.
Vue 3, TypeScript, Pinia and Tailwind on the frontend. FastAPI, PostgreSQL and the Gemini model via google-genai on the backend. Docker deployment on Railway.
A builder gives you a layout. Here, you get a diagnosis: what the offer asks for, what your resume actually covers, what is missing, and what to write instead.
The resume and the offer are sent to Google Gemini for the analysis, then processed in memory on the server. For signed-in accounts, the analysis history can be saved. Only your account (and that history) is stored long-term; you can delete everything from your profile. Details: Privacy page.
Yes. PostHog (EU region) measures product usage and errors. Once you sign in, your account can be linked to events. See the privacy policy.
It is an estimate calibrated on the criteria extracted from the offer, not the exact score of a proprietary ATS. It helps you prioritize your fixes, not predict a hire.
Yes. The repository is public under the MIT License and includes the Dockerfiles, the getting-started guide and the required environment variables.
Issues and pull requests are open. The contributing guide covers the local setup, commit conventions and the review process.
