KAUSHAL DEVRARI

Developed backend APIs using NestJS, improving backend efficiency by 30 percent.

Optimised existing backend APIs for improved performance and reducing response times.

Team: Voice Intelligence - Language AI

Developed SystemApp, AccessibilityApp and MediaResolver Bixby Capsule for Samsung devices which has 10M+ active Users enhancing user experience and accessibility features to the users.

Implemented Cross goal Dynammic Recommendation - CGDR, recommending utterances based on user usage, utilizing knowledge graphs, improving user interaction efficiency by 15 percent.

Contributed to Proactive Bixby, leading to a 20 percent increase in proactive user engagements.

Enhanced LLM Based Bixby, improving conversational accuracy and user satisfaction.

Worked on Anomaly Detection System for Privacy in Samsung Devices using Variational Autoencoders, LSTMS

Team: Voice Intelligence - Language AI

Developed IN-NLU Capsule Project, achieving 96 percent accuracy with a Joint-Intent-Slot NLP Model using BERT.

Utilised ReactJS, ExpressJS, FastAPI for frontend, backend and API development improving system responsiveness by 25 percent.