.Leadership.
The team behind the stack
Eight years building Bangla speech, language and vision models in house, and the operators turning them into products banks and enterprises run on.
.Why this team.
There was no dataset to buy and no base model worth starting from
Building a language model for a language the industry treats as a long tail is not a funding problem. There is no dataset to buy, no benchmark to target, and no pretrained base worth starting from. You need speech scientists who will spend years on acoustics and corpora before there is a product to show anyone.
That is the team that started here in 2018, and it is the reason Ekush LLM, SenseVoice and SenseSpeech are ours outright rather than licensed. It is also the reason we can put those models inside a bank's own data centre, which you cannot offer if the intelligence belongs to somebody else.
The people below span that whole arc: the research that made the models, the engineering that turned them into products, and the operators taking them into banks, telecoms and enterprises across four markets.
The people