training NLP models for spoken languagetasks included:- Text summarization- Keyphrase extraction- Semantic search- Punctuation/truecasing restoration- Disfluency removal- Speaker change detection- Language models for speech recognitionThese various tasks were primarily used to enhance Clova Note, which generates voice-based meeting notes. To achieve this, methodologies such as - task-adaptive fine-tuning of pre-trained language models (e.g., GPT, BERT)- parameter efficient fine-tuning (e.g., LoRA, p-tuning)- hybrid systems between sparse models (TF-IDF, BM25) and neural models were used.framework experienced:- ML training: pytorch, Tensorflow, HuggingFace, Fairseq- ML Inference: Triton, TFLite, Torchscript, Onnx, TensorRT- CI/CD: Docker, Kubernetes, github action- Storage: Hadoop

- Acoustic Analysis group- Project : 1) Multiple Keyword spotting in sentences 2) Deep Learning for Phoneme recognition