Voice AI startup · co-founded · 2024 – 2025

Realty Hybrid RAG + SQL Communication Platform

Co-Founder & CEO

Challenge

Realty brokers waited up to two days for answers that spanned structured listings and unstructured project docs. SQL-only or vector-only retrieval could not handle mixed queries, which drove churn across a developer's portfolio.

Approach

• Built natural-language query system combining structured database (Supabase) and document embeddings (Qdrant).
• Implemented hybrid RAG + SQL routing so agents pick structured vs. unstructured paths based on query intent.
• Deployed as a vertical product for broker and developer workflows.

Tech stack

Supabase · Qdrant · Hybrid RAG + SQL · LLM agents · Python · TypeScript

Outcomes

• Broker query turnaround reduced from 2 days to minutes
• Reduced broker churn across the realty developer's projects
• Reusable hybrid retrieval pattern for other verticals

Context

Realty developer portfolio (customer anonymized): brokers querying listings and project documentation.