Samarth Jangda

Developed a complete system to track all client sessions conducted over Zoom. Leveraged LLMs to analyze recorded sessions across predefined behavioral and conversational parameters. Designed and implemented a performance analytics dashboard. Built an AI-powered chatbot to assist stakeholders. Developed an MVP system for multimodal analysis of Zoom sessions (audio + visual) using Whisper ASR and MFCC-based speaker diarylation. Developed an AI training assistant (MVP) using a Retrieval-Augmented Generation (RAG) pipeline.

Designed and implemented a real-time Automatic Speech Recognition (ASR) system using Hugging Face’s Wav2Vec 2.0. Developed and automated end-to-end training and deployment pipelines using Shell scripting and Docker. Led a small cross-functional team to deliver research-driven projects in Video Emotion Recognition, Body Gesture Analysis, BERT-based NLP systems, SeGAN for Noise Cancellation, and Personality Recognition.

Engineered Encoder–Decoder ASR models with a vocabulary size of 425,000+ tokens. Designed and automated a model development pipeline for live ASR testing. Developed a WebSocket-based automated testing framework to evaluate model performance via Word Error Rate (WER) analysis. Conducted R&D on RASA-based chatbots for domain-specific conversational frameworks.

Designed and developed a Flask + Dialogflow chatbot application for field data collection. Built an internal communication and collaboration platform using Flask and Socket.IO. Contributed to streamlining organizational workflows by introducing automated data handling.