NLP Cloud

We just reached 30 users, which is really cool. We need to keep looking for more users and be reactive about the feedbacks. Maybe there are new killer features that we can easily implement but we never thought about it yet. Maybe we'll need to focus much more on SEO from now on.

Many users were asking for transformer-based models from Hugging Face, in addition to the existing spaCy models.
So we added 4 transformer based models to our API, and we now support the following use case thanks to the best state-of-the-art models available today:

• Named Entity Recognition (NER)
• Text classification
• Text summarization
• Question Answering
• Sentiment analysis
• Part-of-speech tagging

We had feedbacks about the fact that the API was amazing, robust, and intuitive, but that the website and the documentation were pretty ugly. So we refactored both of them and came up with a brand new looking.
We think it's important to have something nice to look at in order to reassure customers.

One of the 10 first free users really loved NLP Cloud, and is an amazing guy. He had some technical questions and we had long and interesting discussions through support. He also gave tons of great feedbacks about NLP Cloud and we did our best to integrate most of them in a couple of days. He seems he was impressed by our velocity and responsiveness :)

I developed the cloud service by myself. I had an hesitation whether I should only include spaCy pre-trained NLP models first, or already offer the possibility to upload one's own custom models too.
I eventually opted for a semi-automated way of uploading spaCy models, and I completely automated it 1 month later when I had a moment, after spending a couple of weeks on the marketing launch first.
I mainly used Reddit, Twitter, and Linkedin for my launch, plus some work on SEO too.
I had my first 10 free users in about 2 weeks.

As a programmer I found it quite hard to deploy my spaCy NLP models to production. I looked for a cloud solution but I did not find any... So I thought it would be nice to create one for both custom and pre-trained spaCy models.
I also thought it would be nice to create a whole annotation/training/testing/deployment workflow for spaCy models in the cloud (as I definitely wanted such a thing in my job). But I realized it would be much more work and I wasn't sure there would really be a market for that, so I dropped this second idea.