Artur Nowakowski, Co-founder and CTO of Laniqo, joins SlatorPod to talk about the language technology platform’s (LTP) origins, business model, and research-driven approach to AI translation.

Artur shares that Laniqo emerged from machine translation research at Adam Mickiewicz University after his team won a WMT 2022 shared task, attracting PONS Langenscheidt, which became the company’s main investor and helped commercialize the university’s technology.

Laniqo initially developed its own neural machine translation models but shifted toward open-source large language models as their translation capabilities improved. According to Artur, controlling and adapting these models remains essential for domain-specific use cases and lower-resource language pairs.

He highlights Laniqo’s work with Central and Eastern European ecommerce platform Allegro, where the LTP supports the translation of hundreds of millions of product offers. Key challenges include scalability, cost control, terminology, limited source context, and detecting critical errors across volumes that human linguists cannot review manually.

Laniqo is also developing quality estimation tools that identify error spans, assign MQM categories, and suggest corrections. Its recently published ForMaT dataset supports research into PDF translation that preserves document layouts without relying on conversion to Microsoft Word.

Looking ahead, Artur outlines how Laniqo plans to expand beyond text translation into voice, images, and broader language AI applications, while continuing to prioritize privacy and deeper domain adaptation.