The AI-assisted Glider Evaluator is an internal tool developed by Aditi Consulting's Engineering Practice to streamline the technical interview process.

What it does

It automatically connects to your active Glider session, retrieves the candidate's answers from a specified assessment, and uses Claude AI to evaluate each response as a Subject Matter Expert (SME).

The evaluation covers the following traits per question:

Correctness Comprehensiveness Expertise Attention to detail Clarity

What it evaluates

Essay, coding, and video questions that require human review. Multiple-choice questions are skipped — they are already auto-graded by Glider.

Output

A detailed score table with justifications per trait, an overall percentage score, and a written summary. Results are automatically written back to Glider and can be exported as a PDF.

Technology

Built on the Anthropic Claude API (claude-sonnet-4-6) and connects to Glider via its DDP WebSocket protocol. No candidate data is stored — everything is processed in real time.

First-time setup (once per browser)

1. On the Step 1 screen, drag the ⚡ Glider Token button to your Chrome bookmarks bar.
2. That's it — you only need to do this once.

Every evaluation

1. Open hire.glider.ai in Chrome and log in with your credentials (including 2FA if prompted).
2. Click the ⚡ Glider Token bookmark you saved. A confirmation message will appear.
3. Come back to this app and click "I'm logged into Glider".
4. Fill in the form:
   
   • Candidate Name — as it appears in Glider
   • Assessment Name — partial name is fine (e.g. "Android Developer")
   • Job Title — the role being evaluated
   • Seniority — sets the evaluation bar
   • SME Role — your area of expertise for grading
5. Click "Evaluate Candidate" and wait ~30 seconds.
6. Review the results. Scores are automatically written back to Glider.
7. Optionally click "Export to PDF" to save the report.

Token expired?

If you see an authentication error, simply go back to Glider, click the ⚡ bookmark again, and retry.

Tips

The candidate name and assessment name don't need to be exact — partial matches work. The app finds the first matching result.