Kepler has engineered a fine-tuned language modeling pipeline that is continuously trained on real-time clinican feedback which we use as model reinforcement. We've created a proprietary approach that "embeds" or "compresses" patient intake history, past notes, and symptomology from the current appointment in order to generate semantic, comprehensive clinical summaries, treatment proposals, and diagnostic insights. Additionally, our dialogue transcription pipeline integrates live speaker identification and utterance detection in order to retrieve rich converstaions between the provider and client. We securely feed the resulting transcription into custom language models to perform summarization and insight retrieval.