Tooploox is a company where we put great importance on our relationships, where we love to learn, and where developing products is not only our job but also our hobby. It was obvious that, after a small pause this year, we needed to organize a hackathon again.
In 2019 we had our last on-site hackathon. Then, as we all remember still too well, we were locked into our home offices for months, but a vision of not seeing each other and not spending time together with the fun we always had while developing our ideas into solutions was terrifying. In 2020 we decided to organize a totally remote hackathon.
The world has changed, and during these few years, we have gained access to so many solutions that support remote work that some of us decided to start working remotely permanently (sometimes from a different country). At the same time, many of us seize all the advantages of hybrid work. According to Gartner, by the end of 2023, 39% of global workers will work hybrid, up from 37% in 2022. This shows that hybrid work is not only a post-pandemic holdover or habit but the way people want to work.
We want to include everybody and be close with every Tooploox employee, no matter how they decide to work. That’s why this year, we organized the first hybrid hackathon.
Tooploox Hackathon 2023 – how did it look this year?
This year 25 people decided to join the hackathon, some of them on-site at the Wrocław office and some of them remotely. They created four teams, which fought for 3 awards in different categories: Best Tech Trendsetter, Best Idea, and Best Execution.
Teams had only 22 hours to prepare presentations of their solutions. And at the end, they won not only glory but also a unique edition to our swag – a bathrobe, which has been the object of the Plooxies’ collective desires for years (all started, of course, from an innocent joke.)
Below you can read all about the solutions the teams worked on during the hackathon.
Tooploox Storage 2.0 – winner of Best Tech Trendsetter
Tooploox Storage is a tool we have already used for a few years. It helps us manage the company’s equipment – assigning it to someone, borrowing, and lending, and ordering new things. We decided to create a new, better tool – with new features and more clean workflows for admins and users. The hackathon was the perfect occasion to take the first step in this project!
At Tooploox, we have the possibility to bid for used office equipment. Making this process more useful was one of the things the team worked on during the hackathon. They designed a flow for creating and participating in auctions and worked on the interface elements to build this flow. The second most significant improvement was to refresh the product base – arranging data from the old base and preparing new data for the latest version. Our goal was to prepare a base in a way that made it possible to view, search, set, and receive notifications in many different contexts.
How we worked
The team consisted of 10 people. Frontend developers, design systems specialists, a UX designer, and the real users – admins of the Storage tool, people who know everything about how it works and how it should work to be perfected.
They used the full range of their skills with the main focus on cooperation and building an understanding between the users and creators: the designers, developers, and engineers.
But most importantly, they spent time together having a lot of fun. They had an opportunity to get to know each other better and talk to each other in a different context than usual.
They plan to continue developing the Tooploox Storage tool. They want to not only make Storage great by creating it as a more helpful tool for admins and users but also to make this project a field where Plooxies can self-develop.
Because of a lack of time, they decided to develop the front end not from UI-ready projects but from mockups. Because of this, they needed to build elements of a design system that would be very easy to modify when they made the transformation from mockups to ready designs.
Another challenge was that they wanted to link Storage with Slack and have it both ways, from Storage to Slack and from Slack to Storage, all in one workflow. Admins set an auction in the Storage tool (added photos, added a description, set the details such as start and end dates, starting price, etc.), and then this auction appears on a dedicated slack channel where Plooxies can bid.
The team said that the essential thought after this project was how awesome and rewarding the cooperation within the team was. They learned how one another works and saw links between their skills and tasks.
The hackathon allowed them to work together on a design system from scratch and see how it looks from different perspectives – such as Product Designer and Frontend developer. The team believes this experience will impact future internal projects in the consistency of elements and ease of creating changes, actualizations, and features.
AI Secretary – winner of Best Execution
At Tooploox, we greatly emphasize taking notes during our tactical meetings. When the person who takes care of these tasks in an AI team heard, during a knowledge-sharing meeting, about the possibilities of the Whisper model, he decided he desperately needed something like this (but better) to support him in writing notes.
The team created a desktop tool to automatically take notes from online meetings in real time and in different languages. The team used an existing solution – The whisper model as shared by OpenAI as open source. The model is great for transcribing recordings but needs to be adapted to write transcriptions in real-time. The team decided adding this feature to the model would be their primary goal. Additionally, they added an option for the tool to write a meeting summary.
How the team worked
The team consisted of 3 people. Adam handled audio preprocessing before the sound was transmitted as input to the model. Karol searched for solutions for mixing signals from the microphone with the signal sent to the speakers. What’s more, he implemented a solution to create summaries. And Szymon created the application’s GUI and integrated it with the backend.
The team used their exquisite knowledge and bench of tools, such as Tkinter, Argos Translate (to translate transcriptions in other languages to English for summarization), Sumy (to create summaries in English), and of course, the Whisper AI model, which, the whole project was based on.
One of the challenges the team met during the hackathon was mixing sounds from different sources. They couldn’t find a solution to program in Python that would work on the various platforms. Ultimately, they used a virtual tool, transmitting a signal dedicated to the speakers to the microphone. This microphone was the source of sound for their application.
The second challenge was speech preprocessing in real-time before being transmitted to the model. It’s not that easy because, at this moment, you need to decide where the statement begins and ends, where you need to cut the signal, and where it should continue. Simple solutions, like monitoring signal strength, do not work in this case because of naturally occurring sounds that are not speech.
The last thing the team found to be a challenge was the possibility of using only offline solutions. The team prioritized data safety and ensuring confidentiality, so they only used offline tools to create their AI Secretary.
The team found speech processing to be a fascinating topic, which brings up a lot of other cases. At the same time, thanks to available libraries (like, for example, flet.dev), you can start to work on beautiful desktop applications very quickly and effectively from the very first.
RescueRoad – winner of Best Idea
The idea for this project came to two of the team members a year ago. Waldek and Michał had a concept to build a system for the Tourist Board of Nepal. They wanted to digitize and simplify the process of licensing and tracking tourists on paths. The idea evolved into a project to rescue tourists who need help on the Nepal mountain paths. Sometime later, Waldek and Michał talked with the Tooploox Founders about this idea. The founders said that they should first think not about the technological aspects, as they are developers, but about business – users, funding, competition, etc. The Hackathon seemed the best time and place to work on it.
The goal of the product itself is to support the rescue process on Nepal’s hiking paths. The idea of the product is to enable faster rescue operations by tracking the positions of tourists. But the team’s goal during the hackathon wasn’t in product development but in the product discovery process.
How the team worked
The team consisted of 6 people: the developers and originators, designers and non-technical specialists, and experts in data and business.
The team’s main focus was to validate the idea – to check the market, the data (a whole bunch of data!), the potential users, the competition, and even the historical facts about tourism in Nepal. And the designers drove the entire validation process using their product discovery skills.
The biggest challenge was to find answers to 4 questions:
- “Does this product make sense?”
- “Who will pay for it?”
- “How will it make money?”
- “Who will use it?”
But this was something that the team expected, and they knew what they needed to deal with to get all the required information before they started to build the product.
The unexpected challenge was a question revealed during the process: “How to build a strong community around this product?” And building a community, in this case, is crucial because it’s not only about earning money and promotion, as only with a big community of users will the app’s full potential be realized. The idea is that RescueRoad tracks tourists’ localizations by using all phones with this app in the mountain range – it remembers when people cross each other’s paths and measures the distance between them in the time from when it passed out of reach of the last signal.
From the first step, the team knew it wouldn’t be a standard hackathon project, where developers have 24 hours to code. They wanted to do this right and not spend time on code that goes into a drawer. The possibility of learning how the beginning of a product’s life should look is the best lesson.
Too many meetings are a struggle, which many people know very well. And while people often don’t even have time to take a break between meetings, they also can’t find time to manage their calendars. This leads to tiredness and a feeling of being overwhelmed. It would be perfect if everybody could have a personal assistant.
The team’s goal was to create, using ChatGPT, a personal assistant that could add, delete and move meetings, add descriptions while sending invites, and also understand the context: see with whom you meet the most often, how long these meetings take, and which can decide which appointments you can resign from.
How the team worked
The team consisted of 8 people: developers, designers, data specialists, Tooploox’s CFO, and two founders. They used their skills to create a most realistic experience – their focus was to build a tool that can help people manage calendars, and not just another tool on which people need to manage their calendars by themselves. What’s more, the tool looks excellent.
The main challenge was integrating Google Calendar with ChatGPT. ChatGPT communicates in natural language, while APIs need some kind of machine language, e.g., JSON. That’s why the team needed to provide well-prepared examples of all supported operations to ChatGPT.
Testing the solution was another challenge the team faced. They needed filled calendars and didn’t want to spend time imitating them. The team decided to trust their assistant and tested it using their own calendars. As a result, many events were canceled incorrectly. Others were left on randomly selected days which have been left as a cheeky reminder of the hackathon.
The goal the team set at the beginning of the hackathon was to prove that ChatGPT can be integrated and used to control different tools. Langchain turned out to be an excellent framework for doing such things with the low effort needed to get a working prototype.