Personalized AI Assistant that Ensures the Privacy of User Data

Self AI

The goal

Develop a personal AI assistant with an advanced level of personalization: a chatbot that helps users choose products and services based on their preferences. Achieve high flexibility and scalability of the project through the collaboration of multiple AI agents.

Timeline

9 months

Year

2024

Technologies

How the Web Service Works

Self AI is a chatbot powered by an LLM that remembers user preferences. The system builds a user profile based on the person's requests.

Self AI analyzes your queries and gradually forms a personalized profile — remembering what you like, where you are, what you're searching for, and what interests you.

This data is used to provide highly relevant and personalized responses, as well as to find suitable information, products, and services online.

You can view and edit your preferences in your personal account — both those detected automatically and those added manually.

For example, if you previously asked where to buy food for a German Shepherd in NYC, then when you ask "Can you recommend a good vet?", Self will provide links to specialists specifically in that city.

Case banner (mobile version)

Monetization Model

The main monetization method of the project is a paid subscription. Depending on the chosen plan, users receive a different number of requests. On the free plan, the number of requests is limited to ten per day.

To enable payments, we integrated with PayPal.

Case banner (mobile version)

Affiliate System

We have set up an affiliate system that provides an additional monetization option for the project. The administrator can add an affiliate to the system — a company that partners with Self and provides referral links to its products and services.

When a user searches for a product, they receive several links to different online stores. One of these links is a referral link leading to our affiliate's website. If the user clicks this link and makes a purchase, the affiliate pays a commission to the service owners.

When adding a new affiliate, the administrator fine-tunes the AI by describing in which cases it should offer the affiliate's links to the user and in which cases it should not.

Case banner (mobile version)

Multi-Agent AI System

We use the Ollama framework, which allows routing requests to both commercial and open-source LLMs. This reduces request costs and enables quick switching between models without compromising response quality.

The system runs multiple AI agents, each with its own role:

  • Classify the query and determine whether an internet search is needed
  • Formulate a search query, send it to Google, and retrieve the results
  • Validate the data from the search engine to ensure it answers the user's request
  • Compose a response for the user based on the information found online and in the database
  • Validate the final response to ensure it meets quality standards

For queries that require web search, we use the Google SERP API.

Calendar Integration

We've integrated the service with Google and Apple Calendars. Users can connect their preferred calendar so that their personal AI assistant is always aware of their plans and tasks.

How can this be used? For example, a business partner from Jakarta suggests a meeting tomorrow at 5:00 PM Indonesian time. You can check your schedule in a second by simply asking the service: "Hi! What are my plans for tomorrow at 5 PM in the Indonesian time zone?"

Case banner (mobile version)

Admin Panel

We've developed an admin panel where the service owner can access financial information about subscriptions and referral link clicks in just a few clicks. This includes detailed data, such as the number of clicks on links from a specific affiliate.

Case banner (mobile version)

Project on Social Media

Jonathan MacDonald, the owner of Self, an Australian entrepreneur and thought leader, is actively sharing the product's development in the media. The project's Telegram chat currently has nearly 9,000 members. Every community member can contribute to the growth and improvement of the AI assistant.

Project team

Zakhar Doroshenko

Project Manager

Danil Danilov

Frontend developer

Vladislav Kiribyatev

DevOps, Backend developer

Ilya Shinen

Data Scientist

Danil Seliverstov

UX/UI designer

Maxim Anokhin

QA

Ready to discuss your project?

Our contacts

Fill out the form to the bottom or email

Email: business@unistory.orgTelegram: unistoryapp

We'll get back to you shortly!

By clicking the button, you consent to the processing of personal data and agree to the privacy policy.

Almaty office

st. Rozybakieva 289/1, office 36,
Almaty, Kazakhstan, 050060

Integrating the future


© 2025 Unistory