Junaid Bhat

Developed a conversation-AI based agent for mental wellness, including Depression Detection (using fusion based deep learning architecture), Valence-Arousal Module (using deep 3d features), Phq-Scoring module, ASR-module, and TTS-module. Actively working on open-source platforms for speech recognition like NEMO and DeepSpeech, providing in-house solutions. Experience includes POCs for speech recognition, speaker diarization, noise suppression, and finetuning/transfer learning of ASR-NEMO on Indic datasets. Built an NLP-based machine learning framework to detect personal/sensitive information (PII) from data streams, calculating severity and masking data accordingly. Created a Healthcare Agent combining multiple AI technologies, including in-house ASR and TTS, a GPT2 based model for script-less conversation, deep learning based emotion detection, and NLP based intent detection. Developed a Machine Learning based Robocall detection system that extracts data from call details records and predicts if an incoming call is RoboCall/Non-Robocall. Developed a POC that generates all possible questions and answers from a given set of text using NLTK libraries and Spacy NER model.