Rushil Shah

Designed the system across realtime AI, telephony, mobile, vision and location-aware context, with a shared product experience rather than disconnected demos.

Built a live location pipeline that keeps the model grounded in a user’s surroundings throughout a conversation.

Reworked the sensing and narration loops to eliminate a 150–200m position lag without making responses feel slower.

Created layered memory for durable preferences, trip history and fresh session context while keeping prompts focused and predictable.

Built toward a single assistant that can explain what is nearby, understand the camera and coordinate the next step of a trip.

Owned the product from architecture and interface design through AI orchestration, rendering infrastructure and private beta.

Engineered a coordinated media pipeline for transcription, shot planning, generated b‑roll, voice and final composition.

Built an asynchronous render system with BullMQ and Redis so computationally heavy jobs remain reliable without blocking the editing experience.

Moved interactive previewing to HTML5 video and canvas for responsive scrubbing while keeping Remotion and FFmpeg for final-quality output.

Designed the editor around intent: creators can shape pacing, captions and visual direction without managing a traditional timeline frame by frame.