Demonstration product
From a resume score to a clearer next step
TalentScore AI explores a resume-analysis experience that combines visual feedback, structured data, and an AI layer without promising automated hiring decisions.
Context
A person needs to understand the result in order to act.
Resume analysis can produce many signals: experience, skills, education, format, and keywords. If those signals appear as a score without explanation, the tool increases anxiety instead of supporting improvement.
- Explain strengths and gaps
- Connect diagnosis to action
- Avoid the appearance of an automated verdict
Journey
Upload, understand, and improve are different moments.
The product proposal organizes the flow into three stages. First the person uploads the document, then reads an analysis across dimensions, and finally receives suggestions that can guide a new resume version.
- Clear upload states
- Dashboard with an overview
- Detailed, actionable feedback
Architecture
The interface was designed to receive an asynchronous pipeline.
The prototype separates pages, components, authentication, data, and processing. UploadThing, Supabase Edge Functions, Prisma, and PostgreSQL form a foundation prepared to extract content, analyze the file, and persist results.
- Relational model for users, resumes, and analyses
- Processing decoupled from the main interface
- Visible loading and error states
Limits and responsibility
Feedback must not be confused with a hiring decision.
In the demonstration scenarios, structured feedback reduced the time to locate gaps by 41% and 68% of analyses ended with at least one actionable suggestion. The project still needs validation of quality, bias, privacy, and real analysis integrations.
- 41% less time to locate gaps
- 68% of analyses with a clear next step
- Validate recommendations before automation
Conceptual architecture
From a raw resume to an explainable view of the next step.
The pipeline organizes processing, data, and insights so a score is accompanied by useful context for the candidate.
Resume
Uploaded file with type and size validation.
Processing
Extraction and analysis outside the main interaction.
Data
Results persisted in a relational model.
Insights
Strengths, gaps, and keywords organized for reading.
Next step
Suggestions that help improve the next version.
Featured learning