Vasu Rastogi

At RocketFrog.AI, I build and maintain voice-enabled AI applications using LiveKit and LLM services, including real-time agent sessions and speech-driven interactions.

I worked solo on the AI front end and back end with a startup to build its product from zero and help secure its first major US government client. I also helped the client secure FedRAMP Moderate (class C) certification.

I started my career by fine-tuning Whisper and wav2vec2 models to run locally for Indic-language ASR. I also debugged and hardened applications for FedRAMP-related deployment requirements.

In selected technical work, I built Python workflows for Whisper-based transcription, voice activity detection, and speaker diarization. I explored speech enhancement and multilingual transcription and translation under constrained compute environments.