HIMKOSH · DATA PLATFORM
Himkosh
Record, transcribe, translate, annotate, and write the data that powers Himalaya models and agents.
Himalaya data layer
From raw data to model-ready datasets.
Himkosh brings human data collection and annotation into the same workflow as Playground, Agents, and Benchmarks.
0
Total records
0
Active datasets
0
Languages
—
QA pass rate
Start a data workflow
Choose how you want to collect or prepare data.
Record
Collect prompted or spontaneous speech with language and capture metadata.
Transcribe
Pair audio with accurate human transcripts for ASR and speech datasets.
Translate
Create multilingual source-target pairs with review and quality checks.
Annotate
Label text, documents, images, and other records using custom schemas.
Write
Collect handwritten Nepali — sentences written on paper or canvas, photographed and cropped.
Datasets
Recent Himkosh projects
No datasets found
Create a dataset to start collecting or labeling data.