research Audio foundation models & voice-to-voice Research at Sarvam AI on audio/speech foundation models, voice-to-voice systems, and music generation. Trans-KBLSTM — knowledge-enhanced tabular reasoning External-knowledge-enhanced Transformer + BiLSTM for tabular reasoning. Best Paper, DeeLIO @ ACL 2022. work Speech & audio engineering Indic ASR/NER, semantic voice-activity detection, and open-source speech tooling.