I built an AI-powered search brain for OmniAxes that understands both text and voice. It uses Gemini to convert free‑form queries into semantic vector embeddings, then performs similarity search across all active posts and products with Firestore’s native vector index returning the most relevant results instantly. The same engine also powers a conversational voice concierge: it transcribes spoken requests via Whisper, decides whether to ask a clarifying question or execute a precise search, and even handles username/profile lookups by stripping extraneous words from the query. Alongside each search, I generate luxury‑concierge verdicts and smart filter chip suggestions to guide users. Rate limiting and App Check protect the endpoints, ensuring the AI remains fast and secure. In short, I built a multi‑modal, context‑aware search system that combines natural language understanding, semantic matching, and voice interaction to make product discovery seamless.