How Tayyari works

The pipeline in one view

1. Sources. A resume, and GitHub, deployed apps and public links when you have them.
2. Readers. One small model reads each source and writes down facts with the lines they came from.
3. Skill graph. The facts become domains and skills, each with a score and its evidence.
4. Next roles. The roles you are closest to, with what you already show and what you lack.
5. Posting match. Paste a job posting and see every requirement against your evidence.
6. Prepare. A plan for a role, built from the gaps in your graph.
7. Practice. Mock interviews that ask about your own work.

Step 1: what you connect

You start with a resume, as a PDF or a Word file. That is enough on its own. Everything else adds evidence and sharpens the graph:

• a GitHub profile, where Tayyari reads your repositories, their READMEs and their tests;
• apps you have deployed, including apps built with Lovable, which it reads from the public page;
• any public link, such as a portfolio, a blog or a Kaggle profile;
• notes you paste in, for example the About section of a LinkedIn profile.

Tayyari does not read LinkedIn pages, because LinkedIn blocks automated reading. It asks you to paste the text instead. It also refuses to open addresses on private networks, so a link cannot be used to reach anything it should not.

Step 2: the readers

Plain code does the first pass. It pulls the text out of PDF and Word files, calls GitHub's public interface and fetches public pages. Then a separate small model reads each source, all at the same time, and extracts skills, areas of work, role hints and facts. Every fact points to the line or passage it was taken from.

A quick first read of your resume runs beside them, so early signals show on screen while the rest is still working.

Step 3: the skill graph

A larger model reads the readers' output together with your resume and builds one graph: four to seven domains, the skills inside each, a score from 0 to 100 for every one, and the evidence behind every score. The score is Tayyari's reading of your sources, not a test result.

Before you see the graph, it is checked. Each claim is compared with the text it cites, including names, numbers and identifiers, and a claim the text does not support is dropped. The skill graph page shows what you get.

Step 4: next roles

The same graph is compared with roles. Each role lists the requirements you already show, the ones you do not, and a rough estimate of how long the gap might take. Roles are ordered by how close they are to you and described in words, never as a percentage. Selecting a requirement opens a conversation about what it means and what to build.

Steps 5 to 7: match, prepare, practise

Match. Paste a job posting. Tayyari reads the posting on its own first, then checks each requirement against your evidence and tells you where you stand in words, never as an "ATS score". It can also build a resume and a cover note from facts you can defend. See job match.

Prepare. For any role, Tayyari turns the gaps in your graph into one next move and a plan sized to the hours you have each week. See career plan.

Practise. Mock interviews ask about your own work and quote your answers back to you. See mock interview.

What code does and what models do

Code computes: scores, weights, selection, totals and the bands in a match. Models read, classify and phrase. Whatever a model writes about you goes into a typed slot and is checked against the facts it cites. A sentence that is not supported is dropped, or replaced by the fact's own wording. Numbers are never invented: if a number would help and is missing, Tayyari asks you for it. Job postings and documents are treated as data, and instructions found inside them are reported, not followed.

Questions