Fix AI App Accuracy Bug

My mobile application relies on an on-device AI model, and users have begun reporting that its predictions have become noticeably inaccurate. The code runs without crashing, yet the results deviate from the expected output, so I need a developer who can focus exclusively on resolving this accuracy issue.

Here is what I need you to handle:
• Inspect the existing inference pipeline, including any preprocessing steps, to locate the root cause of the inaccurate results.
• Correct the bug—whether it lives in the model loading, the feature engineering, or the post-processing logic—so the predictions align with our benchmark data.
• Validate the fix with a set of sample inputs I will provide and document the change so future maintenance is clear.

The project is considered complete once the updated build consistently produces correct results on the provided test set and passes a quick smoke test in both Android and iOS simulators. If you have experience debugging TensorFlow Lite or Core ML implementations inside a Flutter or React Native environment, that will be a plus, though any solid mobile AI debugging background is welcome.

Let me know how soon you can start and your rough estimate of the time required once you’ve reviewed the codebase.