Python Face Recognition System Development

I am seeking an advanced face recognition system developed in Python that can efficiently handle a large-scale dataset of over 10 million images. Each individual in the dataset will be represented by only one image paired with a unique tag. The system will be able to identify whether a new input image matches any existing entries in the database, and if so, provide a similarity percentage.

Key Requirements:
*Single Image Training: The system should be trained with only one image per individual, with no duplicates or multiple images for the same person.
*Accuracy & Performance: It must accurately identify individuals, even in challenging conditions such as:
*Varied lighting
*Partially visible faces (75-80% visibility)
*Dark or overexposed images
*Recognition from older photos (e.g., detecting from a 15 years old image which trained photo has 30 years old man)
*Speed: The system should return results within 3 to 15 seconds
*Database Integration: The trained data must be stored in a database, not in flat files, to ensure faster and more efficient searches as the dataset grows.
*Scalability: The system should be capable of training and recognizing images from a dataset of millions of individuals.

Deliverables:
A Python script that implements the face recognition system, capable of handling millions of entries.
Video tutorials explaining:
*Database setup and how to integrate it with the recognition system.
*Running the script step-by-step.
*Training additional images and updating the model.
*Improving the model (if applicable).

Additional Information:
I am an experienced Python programmer and have worked with face recognition libraries. However, I encountered limitations in my own attempts, especially when scaling to a large dataset. I need a high-level, robust solution that can handle complex scenarios and deliver fast, accurate results.