Build a career profile from evidence

A practical guide to connecting career history, credentials, tasks, and skills using open tools.

This demonstration starts with one person's records. It preserves the difference between what a person reports, what a document supports, what an institution verifies, and what an occupational mapping suggests. The same evidence can support several views of a career without turning every inferred skill into a credential.

Where badges and wallets fit

Digital badges describe achievements; wallets let people store and share credentials. This project adds an analytical layer: connecting descriptions of work to common task identifiers, tracing those tasks to activities and skills, and eventually comparing the profile with job postings. An inferred connection is not an award or an assessment of proficiency.

1EdTech Open Badges and Comprehensive Learner Records support portable achievement records. HR Open Standards' Trusted Career Profile also accommodates self-attested career information. CTDL and CTDL-ASN describe credentials and competencies. These are complementary structures, not interchangeable file formats. A digital signature can establish who issued a record and whether it changed; it does not by itself prove someone's current skill.

1. Assemble a dated career history

Record roles, organizations, education, and credentials separately. Keep overlapping roles and transitions visible. For each fact, retain its source, dates with their original precision, and any correction or unresolved disagreement. A period of doctoral study belongs in the timeline even when no occupational task profile has been assigned to it.

Use your own documents with permission. Keep addresses, private correspondence, transcripts, and unpublished work outside the public website. A public profile can name a source without publishing the document.

2. Record the strength of the evidence

Distinguish personal recollection, a statement in a supplied document, independent corroboration, and issuer verification. A résumé is a source for a claim, not independent confirmation. Record expired credentials as historical achievements; do not infer that they remain active. A certification guide describes requirements, not proof that a person met them.

For example, this profile records the author's reported PSIA Level I certification separately from task candidates extracted from a current PSIA standard. The September 2026 standard is not evidence of the requirements in force when he qualified in the 1990s.

3. Keep two routes to tasks separate

1. Occupation-based inference. Review candidate occupation codes for each career role, then use existing posting-derived task profiles. This suggests activities associated with an occupation; it does not show that the individual performed every activity.
2. Document-based retrieval. Run TaskMatch on selected descriptions of work or credential requirements. Save the exact input, sentence or page locator, task ID, model version, threshold, and output. Keep the source type visible: a requirement in a guide differs from an activity reported in a résumé.

Task retrieval used JAAT 1.1.3 with its task classifier and thenlper/gte-small embeddings, a cosine-similarity threshold of 0.80, and a separate flag for matches at 0.87 or above. The document analysis covers 21 selected Level I statements and 34 work-history lines from two historical résumés.

4. Review the matches before making claims

Read each task against its source sentence. Check the action, the person performing it, and the context. Accept, reject, or leave uncertain, recording a reason. A high similarity score does not replace review.

For example, a criterion about responding positively to feedback matched a task about providing feedback. A skiing statement matched a task about constructing inclines. These illustrate why retrieved candidates remain separate from reviewed personal evidence. Personally confirmed PSIA learning activities are identified separately from other retrieved candidates.

The profile owner can suppress and restore task IDs using Manage task exclusions. Exclusions apply across career perspectives. Retained USAJOBS task weights are normalized within each occupation before recalculating activities and skills; source records remain intact. Choices are saved in the browser and can be downloaded. They do not modify the public website for other visitors.

5. Build transparent scores and alternative views

The profile starts with distinct posting–task pairs from USAJOBS. Task weights are normalized within each occupation, then combined using the selected occupation weights. O*NET links distribute these weights across activities and skills; unmapped weight remains visible.

Show what each score means: it is a share derived from occupational task data, not a personal proficiency rating. The current display does not weight roles by duration, estimate skill decay, or infer continued licensure. Dates are retained so alternative choices can later be evaluated. Document matches have not been merged into these scores.

6. Export records without overstating interoperability

Maintain one source record and map it explicitly to each target standard. Validate exports against the relevant schema and preserve evidence and inference status. An issued badge or learner record also requires an appropriate issuer and achievement evidence; an unsigned JSON file is not a verified credential.

This prototype offers a custom career record and a CTDL-ASN skill-concept fragment. They have not completed external validation. Trusted Career Profile and CLR exports, wallet integration, and signed credential issuance are not implemented. See the architecture and scoring details.

7. Compare career perspectives with jobs

The planned job comparison will let a person select a career perspective—journalism, economics, teaching, or another role—and inspect which tasks connect it to historical USAJOBS postings. Retain posting dates, matched tasks, scores, and source links. Keep education, licensing, citizenship, and other eligibility requirements separate from task similarity. This demonstration does not currently rank individual job postings.

What another researcher needs to repeat it

• A source inventory and permitted copies of inputs, with document versions, dates, and page or sentence locations.
• A dated career ledger with explicit evidence status and reviewed occupation mappings.
• Exact software and model versions, retrieval settings, saved inputs and outputs, and review decisions.
• The public posting-data release, O*NET version and crosswalks, counting rules, and treatment of unmapped tasks.
• Build scripts, export schemas and validation results, plus a public/private file inventory.

The downloadable calculation bundle contains the inputs, mappings, exclusions, and weights needed to reproduce the competency scores. The method page explains how to run the calculation and prepare inputs for another person.