Project and study evaluators

Before you spend two weeks building a portfolio project (or three months on a course), JobEmber has two purpose-built agents that score the idea against your career goals and recruiter signal.

Project evaluator

project-evaluator rates a portfolio-project idea on six dimensions:

• Relevance to your target role.
• Recruiter signal (does this thing actually move the needle in interviews?).
• Time-to-value (can you ship something visible in 2 weeks?).
• Differentiation (does it stand out vs. the typical “TODO list app” portfolio?).
• Personal fit (do you actually want to build it?).
• Maintenance load (will it haunt you in 6 months?).

Output: BUILD or SKIP verdict, plus a 2-week timeline + architecture sketch + success metrics if BUILD.

The agent reads your CV templates + target role + existing portfolio (Settings → My Projects). It won’t re-recommend something you’ve already shipped.

Study evaluator

study-evaluator does the same for courses, certifications, bootcamps, etc. Output: DO IT / SKIP IT / TIMEBOX IT (the “this is worth a weekend, not three months” verdict).

Inputs same as above plus: provider reputation, time commitment, cost.

Adding ideas

Settings → My Projects (or → My Studies) → ”+ Add” — paste a URL or write a sentence describing the idea. The brief-generator agent fills in the details from the URL if there is one (title + description) so you don’t have to.

Then click Evaluate. The agent runs (~10s for project, ~15s for study) and the verdict appears on the card. Re-evaluate if your target role changes.

Why these are agents not heuristics

A heuristic (“does the title contain ‘AI’ or ‘data’?”) would be useless. The signal in this evaluation is contextual — “is THIS project meaningful for THIS user pursuing THIS role” — and that’s what an LLM can read. The agent prompts encode the signal explicitly so the model isn’t guessing what you want.