All projects
Live beta · redesign underwayFounder · product, architecture, and UX

Greenlit.

Industry-aware resume coaching that shows job seekers what is weakening their application—and what to improve next.

The challenge

Most resume tools produce a score or a rewrite without showing candidates what the system actually found. Greenlit needed to make ATS feedback legible, actionable, and trustworthy for people already navigating a stressful job search.

The system

  • Parse the uploaded resume into a structured document model.
  • Classify the candidate's likely industry and apply the relevant scoring rubric.
  • Score structure, evidence, keywords, credentials, clarity, and industry fit deterministically.
  • Use AI for grounded coaching and verified rewrites, then carry approved material into the resume builder.

Important decisions

  • Keep the score itself deterministic so model variance cannot change the baseline diagnosis.
  • Give the coaching layer explicit score bands and harvested resume facts instead of asking it to improvise context.
  • Separate ATS-safe templates from a visual portfolio template whose sharing tradeoffs are stated plainly.

How the work moves

  1. Step 1

    Upload and parse

  2. Step 2

    Classify and score

  3. Step 3

    Explain the gaps

  4. Step 4

    Verify targeted rewrites

  5. Step 5

    Build and export

Current evidence

  • The free scan is live at greenlit.cv and does not require an account.
  • The product includes an industry-aware scoring system, coaching pipeline, and 13 resume templates.
  • A broader profile and product-design makeover is underway; the portfolio labels that work honestly rather than presenting it as finished.

What it taught me

  • AI is more useful when it explains a stable rubric than when it invents the rubric itself.
  • Job-search products earn trust by naming uncertainty and showing evidence, not by overstating precision.
  • Document rendering is product infrastructure: typography, pagination, and export fidelity all affect whether the coaching is usable.

Continue exploring

Return to selected work