In December 2023, we celebrated our 30^th anniversary with an event for our colleagues, clients, and partners focusing on continuous innovation over 30 years. At the event, Yonder had an innovation lab with several booths where some of our enthusiastic professionals showcased innovative ideas. All of these innovations were based on artificial intelligence (AI), and here are some of them.

Mirror, Mirror on The Wall

By Alexandra Bucur & Toni Simu

Discover the magic of Mirror, Mirror on the Wall – your personalized AI companion designed for instant insights into your environment. Fueled by state-of-the-art technologies, including Python, Flask, and JavaScript, this innovative creation by OpenAI transforms your daily interactions into delightful experiences. The Wizard, your AI companion, seamlessly integrates into your routine, offering quick, clever answers with a touch of professionalism and light humor. Whether you’re curious about your immediate surroundings or simply seeking information, Mirror, Mirror on the Wall brings the world to your fingertips.

Explore the seamless communication and intuitive design that powers this AI marvel. The underlying technologies enable effortless interaction with your surroundings, turning every question into a delightful exchange. Step into a realm where artificial intelligence enriches your daily experience, and let Mirror, Mirror on the Wall redefine the way you connect with the world around you.

Here is a demo that allows you to play with the innovation.

The way the demo works:

• You hold something in your hand, wear something particular, do something
• You press the Start recording
• Then you ask a question like “What am I holding in my hand?”, “what is behind me?” “Am I sad?”, or any questions about the photo
• Then press Stop Recording

Click here to play with the demo: https://mirror-mirror-on-the-wall.yonder-makers.com

Copilot

by Răzvan Dragomir

Having your private copilot helping you with your tasks so you can spend your time more efficiently. Thanks to Microsoft, the concept of Copilot is taking off. It can assist you in navigating through Windows settings, generating code with GitHub Copilot, or finding answers in your SharePoint documents. The primary role of a Copilot is to help you with your productivity so that you spend less time looking things up and can take action quickly.

We created two web applications with a Copilot feature to assist you in two use cases: submitting time registration and displaying analytical data in charts and widgets using chat-like assistants.

In other words, you command, and the Copilot executes the actions for you in the web application. The innovation is that we don’t send data to OpenAI to interpret; we ask the LLM to generate code to be executed in the on-premises environment instead. This way, data privacy is fulfilled even if the LLM is public.

We have created two demos to give you better insights into how this actually works.

Analytical Copilot that connects to an existing database and generates charts

 

AI Copilot that automates actions on UI based on human input

 

YonderGPT

By Alex Dreptate

Everyone has heard about ChatGPT and its almost magical capability of responding to prompts. Imagine a custom ChatGPT, specifically trained on information about our company. You can either prompt it by typing in questions or literally by talking to it. It has never been easier to find out details about our history, values, colleagues, and even fun facts. The model will try to keep it all professional, but don’t be surprised by its occasional sense of light humor. YonderGPT has a knowledge base containing information about our company since its inception: number of colleagues and clients that have joined over the years, technologies used, and even some funny stories.

Apart from its vast knowledge base, YonderGPT has the following capabilities:

Web Browsing

If prompted about something that isn’t present in its Knowledge Base, the model can access the internet and scrape data about prompted topics. For example, it can gather information about Yonder’s open careers directly from our website.

DALLE Image Generation

DALLE model can be used to generate AI images. YonderGPT can be asked to imagine how our offices used to look or will look in the future based on its knowledge base and user prompts.

How YonderGPT imagined the office in the next 30 years.

Code Interpreter

This functionality allows YonderGPT to run code for several programming languages, such as Python, JavaScript, and Ruby, and even present the output visually. Below we asked YonderGPT to use the data gathered about open careers on Yonder’s website and plot the roles on a graph along with the Years of Experience Required for each of them. Not only did it generate the python code of the graph, but it also presented the output visually.

 

Plus …. Yonder Hackathon

We will host our annual Yonder Hackathon on Friday, 19, and Saturday, April 20, and kick it off with a dinner on Thursday, April 18. Our hackathon is open to all sorts of innovation, not just AI. Your market and clients may need something completely different, and we should go in that direction.

But our hackathon is a great way to experiment with AI, focusing on the demands of your clients and vertical. It allows you to spend two dedicated days creating a prototype that will be relevant for the (near) future. We have just given you some examples of what is possible and feel free to reach out to us and learn more about this. Get in touch with your Yonder contact person, or send an email to [email protected], and we will make sure it reaches the correct person.

Remus Pereni
CTO Yonder
On behalf of the innovation lab team.

Banner image by Unsplash, Rishabh Dharmani