hey, i'm amirlan.

i'm the co-founder @geko, building training data for human-like voice.

we collect data by co-authoring research with experts like linguists, phoneticians, and voice actors. that lets us automate real-time audio reinforcement learning environment creation, making speech models faster, more natural, and more human-like.

previously, i was a founding engineer @speko (yc s26). before that, i did robotics and research.