Skip to main content
Exam Evaluation suits anyone who collects written answer sheets and has to mark them consistently. Here is how four kinds of customer use it.

​
Schools

The problem: Marking a set of handwritten papers is slow, and two teachers marking the same paper do not always award the same marks. How exam evaluation helps:
  • Upload the question paper once, and the AI reads it into sections, questions, and marks.
  • Every student is scored against the same model answers and the same step marking, so partial credit is awarded the same way each time.
  • Scores can be read and corrected before you publish anything, with a reason recorded for each change. The API accepts these calls from your platform, and your platform decides which of its users may make them.
  • Publish provisional marks for students to check, then publish the final result.
  • Class dashboards showing the class average and the questions the class did worst on are available through CrazyGoldFish, though not as a documented endpoint. Ask us if you want them.

​
Coaching and test-prep institutes

The problem: Students sit mock papers every week and want their marks and their mistakes back quickly. How exam evaluation helps:
  • Set up one exam per mock paper and reuse it for every batch that sits it.
  • Each answer comes back with the marks deducted, the reason, and what to do differently, not just a score.
  • Students see the model answer next to their own answer once provisional results are published.
  • A student who disagrees raises a query, which triggers an AI re-check that either revises the marks or explains why they stand.
  • Tag each answer sheet with your own batch, teacher, and student IDs and list them back by any of those.

​
School software platforms

The problem: Your customers want AI marking inside the product they already use, and you do not want to build it. How exam evaluation helps:
  • Call the REST API from your backend and keep your own screens.
  • If answer sheets already live in your storage, pass a public HTTPS link to the PDF instead of uploading the file again.
  • Register webhooks so your server is told when a paper has been read and when an answer sheet has been scored, instead of polling.
  • Pass your own student IDs on each answer sheet. A roster import from an Excel file also exists, arranged through CrazyGoldFish rather than through the documented API.
  • Or hand a student the answer sheet upload link, a hosted screen, and build nothing at all. It shows the marks as soon as marking finishes, so use it only where that is what you want.

​
Self-study and practice apps

The problem: A student practicing at home has no one to mark their paper. How exam evaluation helps:
  • The student photographs their answer sheet on a phone. An uploaded photograph is checked, and a blurred one is refused with a message asking for a clearer picture. Work sent as a link is not checked, so ask for photographs if you want the check to apply.
  • The AI reads handwriting and describes any diagram the student drew, so diagram questions can be marked.
  • Feedback names the specific gap in each answer, which is what a student needs in order to try again.
  • Generate one upload link per attempt and drop it into your app.

​
What exam evaluation does not do

So you can plan around it:

​
Frequently asked questions

It is built for written answers that need reading and judging. Multiple-choice papers work, since a model answer can be a single option, but you get no benefit from the step marking and feedback that the product is designed around.
Yes. Step marking is what makes this work: a student who uses the right method but reaches the wrong number can still be awarded the method marks. Diagrams drawn by the student are described by the AI so they can be marked too.
Yes. One exam holds as many answer sheets as you need, and each sheet carries your own student, teacher, institute, and parent IDs. Publishing applies to the whole exam, not to individual students.
No. You call REST endpoints and read JSON back. Your developers need to handle background work, which means either checking a status or receiving a webhook.
Whatever your platform chooses to show them from the provisional or final result: their answer as the AI read it, the marks, the model answer, and the feedback. The one screen CrazyGoldFish hosts is the answer sheet upload link, which shows the student their marks as soon as marking finishes.

​
Next steps

Features

What the AI marks and what a teacher controls.

How it works

The full flow from question paper to published result.

Developer guide

Evaluate an exam with the API, step by step.

Assignment Evaluation

The lighter option for homework and daily class work.