At BharatGen, I’m developing Hindi and multilingual ASR systems for Indian languages, including SLAM, a compact Speech-Language Aligned Model for cross-lingual speech recognition. I’m also leading development of Shooktam, a multilingual speaker-adaptive TTS system.
At Sanas, I built a real-time system to detect Mother Tongue Influence and an automated evaluation pipeline for accent translation models. I also improved accent-conversion naturalness and fluency through transfer learning and targeted fine-tuning.
At Kore.ai, I led the ML team for GALE, an enterprise Generative AI platform, and architected its MLOps engine with Ray and Kubernetes. For a major Indian bank, I designed an NLP email-triage pipeline that reduced agent processing time by 67%.
Earlier, I developed speech systems at Gnani.ai and GoVivace, including speech enhancement, language identification, TTS, diarization, and Indic-language ASR. I ranked first in the Multilingual ASR Challenge at Interspeech 2021 and published research at Interspeech 2021.

