News
Polish Startup Migam Wins EEC Challenge With Sign Language AI
Katowice-based Migam, which has spent 13 years building access to Polish Sign Language interpreters, unveiled a working AI model for American Sign Language and is now developing one for Polish Sign Language. The Migam AI project won the 4TECH category of the EEC Startup Challenge 2026.
Contents
Polish company Migam, which has spent over a decade building access to sign language interpreters, announced that its Migam AI project won the 4TECH category at this year's EEC Startup Challenge. The company demonstrated a working model that translates into American Sign Language and said Polish Sign Language is next in line.
From video interpreters to an AI model
Migam has run the OMMI.IO platform for years, letting companies and institutions subscribe to on-demand, live Polish Sign Language interpretation over the internet. The service is meant to help in situations where a deaf person needs to handle official, banking, or medical matters without an interpreter on hand.
Migam AI is the next step, an attempt to replace or supplement human interpreters with an AI model that generates sign language translation directly from text or speech. The company already has a working prototype for American Sign Language (ASL) and is developing an analogous model for Polish Sign Language (PJM, Polski Język Migowy), which differs significantly from spoken Polish in both grammar and vocabulary.
A shortage of interpreters as the starting point
The scale of the problem Migam is trying to solve is considerable. According to figures cited by the company, around 200,000 deaf people live in Poland, while the number of professional sign language interpreters does not exceed 400. That works out to several hundred people needing support per interpreter, which in practice means long waits or no access to translation in urgent situations.
We have around 200,000 Deaf people who need access to sign language interpreters, and there are at most 350-400 professional interpreters - Paweł Potakowski, communications director, Migam
A team with experience on both sides
Migam stresses that more than 40 percent of its team are deaf people, which matters when building a translation model. Without native sign language users involved, it is difficult to produce translations that are natural and grammatically correct, since PJM has its own grammatical structure, independent of spoken and written Polish.
Plans through 2028
The company says its ultimate goal is for Migam AI to autonomously support 30 different sign languages worldwide by 2028. That is an ambitious target, since every country and region has its own distinct sign language, and the resources available for training such models, video recordings with annotated vocabulary and grammar, are far scarcer than for spoken languages.
Significance for the Polish market
For Poland's AI industry, winning the EEC Startup Challenge signals that digital accessibility projects can compete with more headline-grabbing topics like generative language models or coding agents. If Migam actually brings its PJM model to production, it could meaningfully cut waiting times for interpreters at government offices, hospitals, and banks.
The challenge that remains is the quality of machine translation for a language that relies heavily on facial expression, body movement, and the space around the signer, not just hand shapes. The company has not yet said when the PJM model will reach user testing or how its accuracy compares with a human interpreter.
Sources: WNP.pl (wnp.pl)
