Software Developer |... • 11h
3. Phase 2: Factual Knowledge Retrieval (RAG) We treat every property as a unique data silo. Hosts upload PDF guest manuals via the CMS. These are indexed into a dedicated Google Discovery Engine datastore. When a guest asks regarding "WiFi" or "Parking," we query only that listing’s specific datastore. Result: Zero cross-listing contamination. No hallucinations. 4. Phase 3: Language Generation: The retrieved facts are sent to Gemini 2.0 Flash with strict constraints such as: "Use ONLY the provided text. Do not add facts." The model acts purely as a formatter to generate a friendly response.
Software Developer |... • 11h
I built an end-to-end AI concierge for Airbnb & Booking com — without hallucinations. Over the last few weeks, I built a fully automated AI workflow that reads guest manuals, understands messages, and replies safely and accurately at scale. This is
See More
Founder at AuraX • 1y
Recently JECRC College, Jaipur (#1 NEC) invited me as a guest & a jury member at the Fish Tank competition. I saw some exceptionally good startups & was amazed at the ideas & depth of thinking! What surprised me was the immense potential the school
See More
Download the medial app to read full posts, comements and news.