At Lytmus AI, I lead end-to-end AI initiatives and shipped real-time voice agents for AI tutoring, supporting conversations across audio, visual context, and UI interactions. The product grew from 0 to 30K+ users, with 500k+ minutes of voice-first engagement.
I built video pre-processing and context retrieval pipelines for real-time lecture doubt solving, cutting costs by over 60%. I also developed vision pipelines for scientific diagrams that improved accuracy to 98% on the JEE NEET bench.
At Desh AI Labs, I trained high-throughput TTS models across 16 major Indian languages and worked on speech models and AI products for sales call analytics and voice agents. My other work includes a Telugu speech recognition project and freelance backend development.

