A year of projects: what we built in 2022–2023

Over the summer we did something a software house rarely finds time for: we went through our projects and wrote them up. The case study section of this website now describes dozens of projects, some of them going back many years. This post looks only at the last two years, 2022 and 2023, and at the work that was actually on the desks of our teams in Lublin and Graz.

As always, we describe clients by industry and type of solution, not by name. Where a case study is available, we link to it.

Connected devices and the hardware around them

Connected devices are where EMBIQ comes from, and in 2022–2023 they were still at the centre of our work.

• For a start-up building sensor-equipped devices for road infrastructure we have been working on electronics, firmware and software since 2021: from the first demonstrator with an AI camera, through counting, timing and tracking variants, to a version for roadworks (case study).
• For a company in the sawmill industry we built software that processes sensor data to calculate the optimal way to cut a board and minimise losses, and we keep developing it under a framework agreement.
• For a waste management technology company we continued work on the Android app used for waste segregation in smart city projects.

What these projects have in common is that they only prove themselves in the field. A device that works on the desk can behave very differently on a construction site or in a customer's home, so we plan tests in real conditions from the first estimate.

Software for chip makers and electronics companies

A less visible, but important part of our work is software that helps semiconductor and electronics companies develop, test and demonstrate their products.

• A tool that automates testing and quality assurance of wireless solutions, including low-level software (case study).
• Configuration, calibration, debugging and testing software for an active noise cancellation chipset (case study).

The users of this kind of software are engineers, and they are demanding. A good tool has to talk to the hardware reliably, log everything and still be simple enough to use at a lab bench under time pressure.

Web and mobile platforms, design and QA

• A streaming platform for sports fans on mobile devices and smart TVs, which we have been developing since autumn 2022 (case study).
• Long-term maintenance and development of a student housing portal, including keeping it running under heavy load when many users connect at once (case study).
• Frontend work and smart contract integration for a cryptocurrency-based real estate investment platform (case study).
• User interface and user experience design for an AI-based sentiment analysis tool (case study).
• QA support for a retail technology client integrating remote devices across Europe (case study).

Our engineers in other companies' teams

A large share of our work in 2022–2023 was team augmentation. Our developers, testers and project managers joined the teams of other software houses and product companies, in Poland and elsewhere in Europe, often for many months. It is a model that works well when a client already has a strong team and needs specific skills quickly, without a long recruitment process.

Working this way taught us how much of the effort goes into matching: finding the right person for a request, and the right request for a person who has just finished a project. That observation led directly to our first experiment with AI.

Our first prototype built on large language models

This year we built a prototype that uses OpenAI models to work with anonymised professional profiles of programmers and match them with the requirements of projects looking for people (case study). In March we presented the idea of an AI-supported bench exchange at Science Park Graz. Today we use AI internally when we match candidates to incoming requests, and our engineers use AI assistants in their daily work for debugging, code generation and code review, as well as in design, analysis and testing.

In a separate prototype we used AI modelling to analyse geopositioned vehicle and ride data and present it in an analytics dashboard (case study).

Our approach so far is pragmatic. We treat a language model as one component of a system, with clear inputs and outputs, and we pay close attention to what data leaves the client's environment, which is why the profiles in our prototype are anonymised. With the EU still negotiating its AI Act, we also follow how the rules for AI systems will look once they are agreed.

What the list tells us

Looking at the whole period, three things stand out. Most projects combine several disciplines: a device, an app and a backend, or design followed by development and QA. Long-term relationships matter more than one-off builds, and many of the projects above have been running for years. And AI is starting to appear as a tool inside otherwise ordinary systems and processes, not as a separate category.

You can browse all projects in our case studies. If one of them resembles a problem you are working on, get in touch and we will be happy to talk it through.