🤖🍹 Drinkobot: Our AI bartender

For AI Days in Brno in the fall of 2023, we saw an opportunity to showcase AI through a fun project. After discussing with Lidi z Baru, a collective famous for their unique bars, we came up with the idea for “Drinkobot,” a conversational AI chatbot tailored for a bar environment. Lidi z Baru, with iconic venues like Whiskáč and 4pokoje, are the top destinations in Brno for an extraordinary bar experience.

We aimed to add an innovative twist to customer interactions, as even the best bartenders can’t be everywhere at once. Drinkobot assisted during busy times, ensuring every customer felt attended to, providing personalized drink recommendations and entertainment.

Developing the AI bartender

The chatbot we developed acts like a virtual bartender, recommending drinks based on user preferences for drink type (short or long), flavor (sweet or bitter), and specific ingredients. The AI adapts to feedback, refining its suggestions to better suit the user’s taste.

Personalizing the experience

To make interactions more engaging, we programmed the chatbot with distinct personalities. Customers could chat with a bot that mimicked historical figures (Karel IV) or popular/fictional characters (Mistr Yoda, Rambo…), adding an entertaining twist to their experience.

Implementation and engagement

Bars implemented the chatbot via a QR code that customers could scan to start a conversation. This feature not only provided personalized drink recommendations but also served as a fresh form of entertainment, sparking conversations and laughter among guests.

This project was a nice example of cross-departmental collaboration: Data Science developed the AI model, the UI team designed a user-friendly app, and our partner bars contributed their drink recipes. Brno.ai helped amplify our efforts with media support, showcasing our innovative approach to enhancing the bar experience.

The feedback was very positive and customers kept asking for it even after Dny AI event has ended. Will we use it again in the future? Who knows? :)

‍

...

‍

📊 EDVART: Open-source data visualization and reporting library

Effective Data Visualization and Reporting Tool (Edvart) was born from the need to reduce the time spent on creating visualizations during exploratory data analysis (EDA).

At Datamole, we have extensive experience with EDA across various projects. We realized that having a collection of commonly used visualizations ready for any dataset could significantly reduce our workload.

For data analysts and scientists, coding to create data visualizations is a crucial part of EDA for understanding data and discovering insights. Crafting high-quality visualizations requires careful dataset manipulation, selecting the right tools, and frequently adjusting parameters, which often involves a lot of repetitive boilerplate code.

Recognizing the need to streamline this process, our team at Datamole developed Edvart, an open-source library designed to simplify and enhance the visualization creation process.

Development

Edvart started as a simple Jupyter notebook prototype. This allowed us to rapidly test and refine visualizations without the constraints of a rigid structure. The prototype saved us considerable time by reducing the repetitive boilerplate code typically needed in EDA. Its success led us to evolve the notebook into a more structured Python package, designed with an API that supports reusability and customization.

If you are interested to read more detail, check out article by Michal B. here.

Try it yourself!

We decided to open-source Edvart in 2023, aiming to share this very useful tool, our knowledge and hopefully also save some headache to our fellow data analysts.

Open-sourcing Edvart has been rewarding, and we are excited about its future. We encourage you to try Edvart, contribute to its development, or provide feedback to help us improve.

‍

...

‍

📡 Precipitation map of the Czech Republic — hacked radar

‍

Inspired by an article in the Computer magazine, our colleagues applied their technical skills to a cool project — Interactive Map of the Czech Republic showing rainfall status in 72 Czech cities.

They basically hacked the precipitation radar of the Czech Hydrometeorological Institute to build a real-time weather map, providing a clear and immediate visualization of publicly available precipitation data.

Technical details

The interactive map is powered by an ESP32 microcontroller and features 72 WS2812 RGB LEDs, each representing a district city. These LEDs light up based on real-time weather data.

The system runs on a backend developed in Python. Here is a very simplified breakdown of how it works:

• Data collection: A Python script periodically downloads precipitation data from the Czech Hydrometeorological Institute’s radar.
• Data processing: The script processes this data and converts it into a list of RGB values, each corresponding to one of the LEDs on the map.
• Visualization: The LEDs light up to reflect current weather conditions in their respective cities. The map cycles through weather states from 30, 20, and 10 minutes ago, along with the current conditions.

We now have the map in our new and cozy coworking space, guarded by our 🐢 Miluška (CEO Assistant — Turtle Division).

This project is a perfect example of how Datamole supports innovative and fun ideas, driven by the creativity and technical skills of our colleagues, purely for enjoyment.

‍

→ We hope you enjoyed this little excursion into what we do when we don’t code data & AI software for our clients. Stay tuned for more exciting updates and feel free to reach out if you want to join us in our journey!