Senior NLP Engineer, DeepAffects, Mumbai
2019 May — present
1. Implemented DeepAffects' Smart Punctuation API and adapted it for time-constrained (more accurate, 50x faster than DeepPunct) ASR use cases. (Including data collection, dockerization of DL models, client libraries and Kuberentes deployments).
2. Built summarization system for conversational data. (Including data collection, building dialogue act prediction models, clustering, dockerization and kuberentes deployments)
3. Built various analytics and metrics (aspect based key-phrase extraction, call scoring, emotion prediction etc.) for DeepAffects' callspot and calleq.
4. Worked on various parts of ASR pipeline (adapting language models for low-resource domains and domain specific phrases, vad etc.).
5. Adapted DeepSegment for FinSBD 2019 shared task (First in French and Close second in English tasks.).
6. Built DeepAffect's Chapter Detection Api (Including image data genetation for chapter type detection, Text Detection and Recognition algorithms and frame level binary search for chapter detection.)