Our verdict
nexocode has built AI and data systems since 2017, from fraud detection and loan default models to an agentic search platform and an edge vision system for deposit return machines (case studies; company register). It suits a company that wants a small senior team to test one AI idea quickly and then build it, and it is less suited to a buyer that needs a large team or a supplier with growing revenue.
The right buyer
- Companies that want to test an idea before committing. It runs two-day AI Design Sprint workshops and builds a proof of concept in a few weeks before a full build (homepage).
- Procurement and operations teams. Its AI procurement system for Orkla Home and Personal Care cut request-for-quote processing time by up to 85%, from four weeks to four days (case studies).
- Buyers with a vision or edge problem. It built a real-time packaging recognition system for Maas Loop’s deposit return machines, from data set to deployment (case studies).
Reasons to look elsewhere
- Its revenue fell in 2024 and 2025 and it reported net losses in both years.
- With a small team, it can run only a few projects at once.
nexocode, At A High Level
nexocode was registered in Kraków in July 2017 by Radosław Kamiński, who is president of its board, a shareholder, and its listed beneficial owner (company register). It started as a classic software house, lost its main client within two years, and in 2018 began turning itself into a “teal” organization with a flat structure, open finances, and self-set salaries (founder story). Its team includes machine learning engineers, data scientists, software engineers, and advisors in life sciences and telecom (about page).
- Registered at Wadowicka 7 in Kraków, Poland (company register).
- Revenue of 3.32 million zloty (about $900,000) and a net loss of 201,099 zloty in 2025, down 14% from 2024; revenue peaked at 6.12 million zloty in 2020 (filed accounts).
- Services cover AI consulting, generative AI development, data engineering, cloud development, computer vision, natural language processing, predictive analytics, recommendation systems, and dynamic pricing (homepage).
- Published cases include Recordya, an agentic enterprise search platform, a group travel CRM for Time4Travel, a data hub for Boussias Group with more than 3 million records, and fraud detection models for insurers and lenders (case studies).
Scores by criterion
We scored nexocode from its homepage, about page, case studies, and founder’s writing, and from its filed accounts in the Polish company register. Technical depth and platform independence score highest; stability scores lowest because revenue has fallen two years running.
| Criterion | Weight | Score | Basis |
|---|---|---|---|
| Production track record | 20% | 6.0 | Published production cases for Orkla, Maas Loop, Recordya, and Boussias. |
| Technical depth | 15% | 6.5 | Machine learning, generative and agentic AI, computer vision, and data engineering. |
| Industry depth | 10% | 6.0 | Retail, travel, finance, insurance, logistics, and pharma cases. |
| Delivery model and handoff | 15% | 6.0 | Design sprint, proof of concept, then build with MLOps. |
| Commercial clarity | 10% | 5.5 | Fixed starting formats such as a two-day sprint; no published rates. |
| Platform independence | 10% | 6.5 | Custom models; Google Cloud partner. |
| Stability and scale | 10% | 4.5 | Founded 2017; about $900,000 revenue and losses in 2024 and 2025. |
| Independent evidence | 10% | 5.5 | Attributed client quotes from Google, Katana, and Leavetown. |
| Overall | 100% | 5.9 | Weighted average |
Where nexocode fits
nexocode is used by mid-sized companies and startups in Europe and North America that want a small AI team to run discovery, build a proof of concept, and then take one AI system to production. On price it sits at the lower end of European AI boutiques, reflecting Polish costs and a small team. Competitors on this site include deepsense.ai for Polish machine learning work, DataRoot Labs for small AI teams, and MindTitan for compact machine learning consultancies.
| Need | Fit | Why |
|---|---|---|
| An AI proof of concept in a few weeks | Strong | Its Design Sprint and Proof of AI formats are built for this. |
| Generative AI for documents or procurement | Good | It cut Orkla’s RFQ cycle from four weeks to four days. |
| Computer vision on edge devices | Good | It built packaging recognition for deposit return machines. |
| A multi-team enterprise AI program | Weak | Its team is small and revenue has been falling. |
What clients and leaders say
nexocode’s leaders and clients describe its work in published announcements. One buyer concern is written from public facts about the firm.
“The support we received from nexocode team was top-notch. They went above and beyond to assist us at every step, providing valuable insights and guidance whenever needed.”
“The software development significantly improved the onboarding processes and ensured that around 10 times more applications could be processed weekly.”
“Nexocode writes strong, clean code while maintaining excellent communication.”
“Their engineers moved from workshop to a working prototype quickly, but the team is small, so we asked who would stay on the project for the full build.”
Where customer reviews couldn’t be found, our analysts attempted to give a voice to buyer sentiment, drawing on real information from review sites, forums and product documentation. Based on: nexocode’s filed accounts and the team list on its about page.
Who leads nexocode
Radosław (Radek) Kamiński founded nexocode in 2017 and leads it as president of the board, a role the firm calls company lead (company register; about page).
