Neev
SOTA pipeline for clinical note / transcription generation for India's dynamic healthcare ecosystem
Links
Date
2024–25
Pipeline vs Whisper v3 Large
Below is a real-world sample comparing OpenAI's Whisper v3 Large against this pipeline.
Hindi, Telugu, Kannada, Tamil mix
Whisper v3 Large
ಇವು ವಿವೆಕ್ ಡೋರ್ ಪೆ ಫೂಡಾಗಿಯ ಸೂಟ್ರ ಆಡೆ ಉಂದಿ ಬೇಲ್ಲಕೊಚಿನೆಡ ಸೂಟ್ರ ಅವರತ್ತರ ಕಿಯು ತೋಮಮ್ಬಟು ಕೆಲ್ಗಡೆನ್ನು ತೋಮಮ್ಬಾಂತೆವಿದು ಅವನುಗೆ ಕೆಲ್ಗಡನೆ ರೆಡಿಯಾಗೆ ಅವಂ ಸೋಲ್ಲಾ ಅಂಗೆ ಇಲ್ಲಾ ಇನ್ನು ಸೋಲ್ಲಾ ಚೆಕ್ಕುಂಟಿ
Neev Pipeline
विवेक, डोर पे फूड आ गया है। చూడు రాడే ఉంది, బెల్ కొట్టినాడు చూడు। ಆ ಮತ್ತೆ ಕೀ ತಗೊಂಡು ಬಿಟ್ಟು ಕೆಳಗಡೆಯಿಂದ ತಗೊಂಡು ಬಾ ಅಂತ ಹೇಳು ಅವನಿಗೆ. ಫುಡ್ ಕೆಳಗಡೆನೇ ಇದೆ ರೆಡಿ ಆಗಿ ಇದೆ। அவன் சொல்லுப்பா, அங்கே இல்லைன்னு சொல்லு. செக் பண்ணு।
Advised on AI strategy and built one of the most advanced clinical transcription and note generation pipelines in the Asia market, tailored for India's complex healthcare environment.
After benchmarking existing solutions, I identified critical gaps in handling real-world clinical scenarios:
- Multilingual conversations — doctor, patient, and attendant each speaking different languages
- Code switching across multiple languages and dialects mid-conversation
- Accurate transcription of medications that are hard to pronounce (e.g., Trastuzumab, Adalimumab, Bevacizumab)
- Noisy audio conditions — sneezing, coughing, patients struggling to speak due to illness
Designed a custom pipeline supporting all major foreign and Indian regional languages with a strong focus on latency, processing 60-minute consultations in under 40 seconds while maintaining superior accuracy.
Helped with implementation of this technology in hospital chains and their EMR systems.