Poseidon

Bengali speech spanning single-speaker narration and two-speaker conversations, so models see natural turn-taking as well as clean prompts. Audio ships as mono 44.1 kHz WAV, and every clip carries a ground-truth transcript with speaker structure and format metadata.

Multi-modal training data collected with consent and licensed for commercial use: speech, real-world video, 3D capture, and reasoning pairs.

Hindi speech recorded through a moderated contributor app, keeping capture quality and formatting consistent across speakers. Audio ships as mono 48 kHz WAV, and each clip pairs the recording with a validated transcript plus duration, codec, and speaker metadata.

Numo is Poseidon's data collection platform: skilled contributors in homes, factories, and workplaces, validated capture hardware, and quality control on every clip. Pick up an open task or scope a custom collection.

Latest updates, announcements, and insights from the Poseidon team.

The Indic Speech release provides 3,845+ hours of transcribed Hindi, Telugu, Tamil, and Bengali speech data for ASR training, evaluation, and error analysis.

An ASR evaluation benchmark for low-resource languages, real-world audio, and production failure modes.

Open-source multilingual ASR evaluation: benchmark WER, CER, semantic similarity, and demographic failure modes in a single run.

The company develops the platform contributors use, the pipeline that evaluates their submissions, and the benchmarks that validate the datasets.

Poseidon collects real world data and refines it into qualified, model-ready datasets for the labs building the next wave of robotics.