From a first interest to a clear next step: add information, understand its connections and control your data.
01Start with interests and a goal
Choose “Build your profile” on the home page. Read the purpose and consent if you want to begin.
Choose your next goal and primary career interest. “Undecided” is a valid choice.
Open the optional preferences section to add work and learning preferences.
Select “Explore my network” to see the effect of your choices. A transcript is optional.
02Add evidence and review it
Choose “Evidence — optional” or “Start with your transcript”. You can try the synthetic record first.
Upload a supported file and follow the on-screen instructions. Extraction can make mistakes, especially with images.
Check each course name, code and grade against the original. Resolve missing values and duplicates before approval.
Approval means you reviewed the information; it does not authenticate the record with your university.
03Understand the connections
Open “My network & guide” and choose a pathway.
Select a question about evidence, interests, people or a learning step.
Select a node to inspect its links. Expand “Inspect path relationships and sources”.
Distinguish student declarations, reviewed evidence, catalog relationships and rule-derived conclusions.
04Turn understanding into an action
Choose a question in the decision guide, then open the development plan.
Read the reasons, evidence limits and gaps before choosing a pathway or training.
Download the development plan for your own follow-up.
Check the original provider’s requirements. Current jobs and training are reference profiles, not bookable opportunities.
05Save your profile and return
Open Privacy and choose how you want to save.
Device storage is optional; avoid it on shared devices.
Export an encrypted backup and import it later. Keep its password; there is no password recovery service.
Withdraw consent or delete local profile data. Downloaded copies must be deleted separately by you.
06Explore knowledge and SPARQL — optional
Open the knowledge explorer from the top navigation without creating a personal profile.
Search for a capability, interest or pathway, then follow its connections.
In the SPARQL lab, select reference knowledge or the synthetic student, choose a question and run it.
Personal queries require separate session permission. Queries are local and read-only, limited to 100 rows and 12 seconds.
07When a result looks wrong
Try the synthetic example to separate input problems from navigation.
Check course codes, grades, review approval and the relevant consent.
For query errors, try a built-in question against reference knowledge.
Save an encrypted backup before refreshing or clearing browser data. Share reproduction steps without your transcript or personal information when reporting a problem.
Ready to explore Kamin?
Start with the synthetic example or explore reference knowledge without personal information.
Unsupported file: use PDF, an image or TXT, or enter courses manually.
Inaccurate extraction: try a clearer, flat image. Check every code and grade and remove incorrect rows before approval. You can start with interests without a transcript.
Saving blocked or storage full: keep this page open. Export an encrypted backup in Privacy & control, free browser storage, then retry saving. Private browsing or device policies may block storage.
Corrupt backup or wrong passphrase: keep the original file and try the correct backup and its passphrase. There is no encryption-passphrase reset service.
Lost device or cleared browser storage: restore a previously exported encrypted backup. Without a backup, the Kamin team cannot recover the profile; you can start again.
Review or contest a recommendation
On a pathway or course card, or in the decision guide, select “Review this judgment” or “Contest this judgment”. Inspect requirements, evidence and source; check your goal and the recommendation’s limits.
Choose a decision, reason and confidence, and enter a name or alias for local review. Acceptance and rejection require all three checks; contest and unresolved remain available without them. Timing is optional and local.
Save locally. Withdraw your decision on the card or delete reviews in Privacy. The activity view retains history, and encrypted backups include reviews; restoration or changed inputs require fresh confirmation.
To hand a potential issue to the team, download the contest’s test candidate. It excludes names, notes, grades, inputs and profile fingerprints. Nothing is sent automatically; the team must confirm a defect with a synthetic example before adding a regression check. Capability-sharing links do not include these reviews.
The trust level of every field
Every row extracted from your document carries a “from the document” badge. If you edit a code, name or grade, the badge becomes “edited after extraction”, the extracted value shows on hover, and the row returns to its state when you restore the value.
Raising a grade after extraction is named explicitly on the review screen. It does not block approval, but the resulting capability is shown as self-declared, and that level travels to skill cards, the shared snapshot and exports.
When you upload a PDF or image we compute local integrity signals: whether a digital signature exists and still covers the file, how many times the file was saved after creation, the producing software, the modification date, and any QR code with a “verify at the source” link. These are signals, not verdicts, and change no result.
From “what evidence am I missing?” or from My skills, attach a project, certificate or internship to a gap. It is recorded as self-declared applied evidence awaiting review, adds no capability and changes no Fit.
From Privacy, download an unsigned portable export in the CLR 2.0 shape carrying the same provenance levels. It is not a verified credential; an institutional system can read it and replace the issuer and proof later.
Task, then evidence, then review
After approval your dashboard opens with three answers: my current goal, my most important missing evidence, and what I do now. Advanced tools remain available but secondary.
Each gap has one executable task on synthetic training data, with steps, a deliverable, a rubric and a review scope. Statuses: not started → in progress → output attached → needs improvement / reviewed locally. Progress is saved and reopened.
Completing the task, evaluating the output and issuer verification are three different things shown separately. A peer or local reviewer opinion stays self-declared and never raises the evidence level, whatever role is typed.
In Matches, compare two or three pathways: evidence present, evidence missing, work required, and your declared constraints if you choose to include them. No readiness percentage.
When describing a project, press “Suggest capabilities from the description”: the local assistant proposes capabilities and quotes the supporting sentence, or abstains without one. You decide what is saved.
From Privacy, import a signed credential to check it locally: structure, issuer against the trusted-issuer registry, signature, validity and revocation status. The registry is empty in this release, so every import stays unverified.