Getting started

Your CV is the starting point for everything. Once uploaded, the system parses it into a structured profile — work history, skills, education — which the score engine uses to evaluate fit. You can have multiple CVs but only one is "active" at a time; the active one is matched against new jobs.

• •Go to the Profiles page and choose "File" (PDF/DOC/DOCX) or "Paste text" to enter your CV as plain text.
• •Optionally add a title ("Senior Backend Engineer") and a vibe ("open to contract, remote only") to give context to the scorer.
• •After upload you're taken to the profile detail page automatically.

Parsing extracts your structured profile from the raw document. This is a one-time step (unless you upload a new version). The scoring engine needs the structured data — it can't produce a useful breakdown from the raw PDF text.

• •After upload the status badge shows "Parsing..." — this takes about 5–6 minutes.
• •You'll see "Parsed" (green badge) on the Profiles page when it's ready.
• •Click "View Profile" to review what was extracted — name, skills, work history, languages.
• •If it fails, click "Reparse" to retry.

The Jobs page shows all jobs in the system. Each job has a structured card built by the JD parser — sector, required skills, seniority level, employment type. You can filter by type (permanent/contract) and arrangement (remote/on-site) to narrow down to roles that match your preferences before scoring.

• •Go to the Jobs page and browse or search the listing.
• •Click a job row to open the detail page: the JD card shows must-have skills, nice-to-haves, what the role actually needs, and any recruiter-flagged quality issues.
• •Use the filter chips (employment type, arrangement, source) to narrow the list.

Scoring runs the LLM pipeline against your active CV and a specific job. The result is a 0–100 fit score plus a full breakdown showing what you match, what's missing, and by how much. You can trigger scoring for one job at a time ("Check Fit") or let the system find your best matches across all jobs.

• •On the Jobs page, click "Check Fit" on any parsed job row to score your active CV against that job.
• •Alternatively, go to the Matches page and click "Find best matches" to score your CV against all available jobs at once.
• •Scoring takes about 5–6 minutes. The Matches page shows an "In progress" badge while it runs.

The score detail panel is the main deliverable. It tells you which requirements you meet, which you're missing, and how the overall score was calculated. The floor/current/ceiling range shows your worst-case and best-case fit given skill overlaps and gaps. Remediation cards highlight the gaps that, if addressed, would move the score the most.

• •On the Matches page, click the coloured score chip to open the full breakdown.
• •Review the Strengths section: each strength shows the skill, how strongly it matches, and the evidence from your CV.
• •Review the Gaps section: each gap shows the missing requirement, its severity, and a suggested fix (where available).
• •The Requirements Coverage table maps every JD requirement to your CV — met, partial, or missing.
• •The score audit at the bottom shows how each dimension (technical, seniority, logistics) contributed to the final number.

Your talent pool lives on the Candidates page. Every CV you upload here becomes a candidate record that can be parsed, scored, and compared across all your jobs. Bulk upload lets you drop multiple files at once — drag them onto the widget or use the file picker.

• •Click "Add Candidate" at the top right of the Candidates page.
• •Drag one or more PDF/DOC/DOCX files onto the upload widget, or click to open the file picker.
• •Each file creates a separate candidate row. No limit on the number of files in one batch.

Parsing extracts structured data from the raw document — work history, skills, education, seniority level. The score engine needs this structured profile; it can't score an unparsed CV. Each CV must be parsed once; re-parse only if the candidate sends an updated version.

• •Find the candidate row in the table. If the status badge reads "Not parsed", click the "Parse CV" button in that row.
• •The badge changes to "Parsing..." while the worker runs — expect 5–6 minutes.
• •When the badge shows "Parsed" (green), the CV is ready for matching.
• •If the badge shows "Parse failed", click "Retry parse" to requeue it.

Jobs are the other side of the match. You can paste a raw JD (job description) directly or import by URL. The system deduplicates by URL so importing the same listing twice is safe.

• •Go to the Jobs page and click "Paste JD" to paste the description directly, or "Import URL" to pull it from a job board link.
• •Give the job a title if the parser can't extract one automatically.
• •The job row appears immediately; parsing begins in the background.

The JD parser builds a structured card — sector, must-have skills, nice-to-haves, seniority expectations, and a recruiter-facing quality assessment. This card drives both the scoring prompt and the candidate-facing preview. Like CV parsing, it takes 5–6 minutes.

• •The status badge on the job row shows "Parsing..." while in progress.
• •Once the badge shows "Parsed", the job is ready to run matches against.
• •You can view the card before matching by clicking the job row to open the detail page.

Matching runs the LLM scoring pipeline for every (CV, job) pair you trigger. You can score one candidate across all your jobs, or score all candidates against a specific job from the Pipelines view. Results include a fit score, strengths, gaps, and a full breakdown.

• •Per-candidate: on the Candidates page, click "Run match" in a candidate row. This queues that CV against all your parsed jobs.
• •Per-job: go to Pipelines, pick a job, and click "Run match against my candidates" to score every parsed CV against that job.
• •Scoring takes 5–6 minutes per pair. The Matches page shows progress with an "In progress" status badge.
• •Once done, click the score chip (the coloured number badge) to open the full breakdown: strengths, gaps, requirements coverage, and a score audit.