Emil Sedgh is chief technology officer of Rechat and an architect of Lucy, its AI assistant for real-estate professionals. He builds software that turns brokerage data, property listings, marketing tools, and customer relationships into automated workflows that can be evaluated for reliability.
Sedgh began in open-source software communities, contributing to KDE and participating in the Tehran Linux Users Group. He mentored a KDE Google Code-in project and helped organize a joint KDE and Debian release event in Tehran. He worked with founder Shayan Hamidi during Rechat’s early development, building a platform that integrated contact management, listings, transactions, and marketing.
In 2023, Sedgh helped introduce Lucy as a way for agents to ask questions about real-estate transaction forms. Its ambitions quickly expanded: Rechat’s existing customer data and internal APIs made it possible to build an assistant capable of executing tasks inside the software agents already used. An early GPT-3.5-and-React prototype demonstrated the potential but was slow, inconsistent, and vulnerable to regressions whenever prompts changed.
- Domain-specific LLM evaluation: Working with Hamel Husain, Sedgh helped move Lucy beyond informal testing by introducing checks based on actual real-estate workflows, observed failure modes, synthetic agent requests, and continuous integration. Their AI Engineer World’s Fair appearance shows how evaluation made improvements and regressions measurable.
- Fine-tuning for complex agent workflows: Sedgh found that prompting alone could not reliably combine conversational answers with structured interface elements, request missing information, and coordinate multiple tools. One workflow finds property listings, chooses the most expensive, builds a website, creates social content, prepares an email, and schedules follow-up.
- Production responsiveness: Sedgh treats latency as a product constraint: decomposing requests into additional model calls can improve execution while making an interactive assistant too slow for working agents.
Named a 2026 Real Estate Newsmaker, Sedgh has built his career around making industry-specific AI accountable to the operational demands of real-estate work.
