Deana.AI

Email-First AI Virtual Assistant

Client
Client
Deana.AI® is a virtual assistant that turns your emails into a convenient productivity tool. A text-based artificial intelligence reads and executes text commands using GTP-3 and a custom NLP model. Users can interact with Deana.AI® by email, messengers, or online chatbots.
Location
Location
United States
Industry
Industry
AI, Startup
Services
Services
Technologies
Technologies
Databases, Machine Learning, Artificial Intelligence, Python, Node.js, React, MongoDB

Product

The idea behind Deana.AI® was to use artificial intelligence (AI) to expand the capabilities of traditional email communication and planning.

For example, if you send Deana an email that says, “lunch with Dave at 5 pm," the AI will update your schedule and send the invitation to Dave. It will also send you a reminder before the meeting. After lunch, you can email the image of your bill, and the system will automatically recognize the items and amounts to include them in your expense report.

Sending the commands takes a few seconds, but it saves a ton of manual work.

Key Features

  • Schedule meetings, events, and tasks
  • Track expenses, bills, invoices, and other financial operations
  • Manage local files and access data in the cloud
  • Search information online (news, articles, PDFs)
  • Navigate using online maps
  • Get short summaries of multi-page texts that you don’t have time to read

Technologies behind Deana.AI® 

Deana.AI® uses natural language understanding, machine learning, and predictive analytics to process user commands, provide a personalized conversational experience, and learn each user’s behavior patterns.

We chose between multiple NLP models to find the best match for our tech and product requirements. After in-depth research, we decided on GPT-3 (Generative Pre-trained Transformer 3), which performed the best during our tests.

Our specialists fine-tuned the GPT-3 model to better operate in our knowledge domain. The model was written in Python and Node.js and run on Nvidia's GeForce 3000 series GPU.

It required both automatic and manual correction to minimize errors and improve operations. Our data specialists analyzed the answers generated by the model and found and corrected errors.

Team

Since Intersog already has vast experience in AI development, we formed the team quickly. As of today, it consists of one each of the following:

  • Backend AI developer
  • Frontend engineer
  • Data scientist
  • UI/UX designer
  • QA specialist
  • PM

Similar Cases

How can we help you?