👉 Before it’s everywhere, it’s in FWDstart. Subscribe for $99 and save 17% on the first year to support independent reporting on the MENA tech and startup ecosystem
Abu Dhabi-based CNTXT AI is opening its first US office as it reports $300 million in “committed revenues”, most of it from American cloud computing groups, AI developers and large enterprises.
The company plans to build a commercial team of 15 to 25 people in Milpitas, California, during its first year, including senior revenue and marketing executives, chief executive Mohammad Abu Sheikh told The Wall Street Journal.
The expansion follows a $60 million Series A in June, co-led by AI71 and BlueFive Capital, and puts CNTXT closer to a customer base already concentrated stateside.
Abu Sheikh said 90% of its clients were in Silicon Valley, predominantly hyperscale cloud companies and robotics labs, making a presence in their time zone increasingly important.
Founded by Mohammad and Hassan Abu Sheikh, CNTXT develops Arabic AI models and supplies the data other companies use to train and evaluate their own systems. Its business spans voice software for Arabic-speaking customers and recordings of physical work used to teach robots how to operate outside controlled laboratory settings.
The company spent two years developing its Arabic-first models. Its Munsit voice platform now has more than 300,000 users, according to figures provided to the WSJ, while its speech-to-text and text-to-speech models have been trained on more than 110,000 hours of Arabic audio covering over 25 dialects. US businesses use the technology to serve Arabic-speaking customers in their own language and dialect.
CNTXT strengthened that business in June when it acquired Actualize, whose Arabic voice agents can handle bookings, transactions and other workflow tasks. Actualize’s co-founders, Muhammed Shabreen and Khalid Ghiboub, joined as CTO and vice-president of AI models respectively. Financial terms weren’t disclosed.
Its robotics work addresses a different data problem. Abu Sheikh argues that systems trained mainly in laboratories can struggle with ordinary tasks, such as folding clothes or picking up objects, despite being capable of more spectacular demonstrations. CNTXT operates robotics training facilities and a collection network intended to give developers more examples of how people perform those tasks in working environments.
The company’s data collection products include first-person video with hands, objects and actions labelled for training, alongside footage from warehouses, kitchens, hotels and hospitals. Its KLKT app pays contributors for accepted recordings of everyday work, which CNTXT says are collected with consent and reviewed by people before labelling.
Abu Sheikh sees the Middle East’s concentration of service and industrial workers, and its modern hospitality and logistics facilities, as an advantage in supplying that material. The ambition is also to bring the resulting tech back to the places where the data originated, with US customers buying recordings to train their systems, while CNTXT works with them to then deploy robots in those same regional environments.
“So, it’s a complete partnership cycle that we are both working towards together,” he told the WSJ.
👉 Before it’s everywhere, it’s in FWDstart. Subscribe for $99 and save 17% on the first year to support independent reporting on the MENA tech and startup ecosystem
