Foundational language AI for Northeast India's indigenous languages
Website · Papers · HuggingFace
MWire Labs builds and ships language models, speech systems, and agent infrastructure for low-resource Northeast Indian languages — Khasi, Garo, Kokborok, Mizo, Nyishi, Meitei, Bodo, Assamese, Pnar, Nagamese, Wancho, and more.
- Language Models — KhasiBERT, GaroBERT, MizoBERT, KokborokBERT, NyishiBERT, NE-BERT
- Speech — NE-MultiSpeech, Garo ASR, NE-TTS (Khasi)
- Translation — GaroNMT, KokborokMT, WMT-winning KrenTransV0-1 (11 NE language pairs)
- Agents & Tools — NE-Agent — multilingual CLI agent for NE languages
- Vision & OCR — NE-CLIP, NE-OCR (Assamese, Garo, Mizo, Nyishi)
- Embeddings — NE-Embed (LaBSE-based)
Published at EACL, LREC, ACL, EMNLP.
All models and code released under CC-BY-4.0 unless noted otherwise.
Shillong, Meghalaya · India