Namy AI

Frequently Asked Questions

How namy.ai works?

For those delving into the website''s about section, an understanding of how namy.ai functions might be of interest. Here''s a concise overview:

I gather data from a substantial number of websites, approximately 10 million, extracting their domain names, contents, and other pertinent high-level features.

The subsequent steps involve extensive cleaning and data processing to form a robust dataset, which, incidentally, consumed a significant amount of time in the process.

Following this, I employ a sizable PyTorch model for training on the dataset. The objective is to generate novel domain names that have not been seen before. Initially, the model predominantly consists of LSTM layers, but I''ve since transitioned to a custom transformer architecture, yielding improved performance albeit with a longer training duration.

Post-training, the enjoyable phase begins: the generation of new domain names. This process involves a level of complexity, as I fine-tune parameters and incorporate specific rules. Essentially, it entails prompting the model to predict new sequences.

The AI-generated domains undergo availability checks using the namecheap API (and additional ones if quota limits are reached).

Subsequently, these domains are temporarily stored in a PostgreSQL database. The Bootstrap frontend regularly fetches data from this database at intervals of approximately one second until the results are ready for display.