DubaiNest AI — a RAG-powered real estate assistant for Dubai

🏙️ I built a live AI product in 3 days. Here's what it does and how it works.

Introducing DubaiNest AI — a RAG-powered real estate assistant for Dubai, built by own from scratch during the AI Accelerator Bootcamp learning by Decoding Data Science.

The problem it solves:

Every expat in Dubai knows this frustration — scattered rental prices, confusing RERA laws, no single place to get a straight answer. DubaiNest AI changes that.

You can ask it:

🔹 "What is the average rent for a 1BR in JVC?"

🔹 "Can my landlord increase rent by 20%?"

🔹 "What is the total move-in cost for an AED 90,000 flat?"

🔹 "Which areas suit a young professional?"

And it answers accurately — grounded in real data, no hallucination.

The tech stack:

⚙️ LlamaIndex — RAG pipeline & query engine

📦 Pinecone — cloud vector database (1536-dim embeddings)

🤖 OpenAI GPT-4o-mini — LLM (temperature=0, factual answers)

🌐 Flask + Waitress — production API server

🐳 Docker — containerised deployment

🤗 HuggingFace Spaces — live hosting, single URL

What I learned building this:

✅ Data quality matters more than model choice

✅ LlamaIndex's {context_str}/{query_str} != LangChain's {context}/{question} — a small difference that breaks everything

✅ Shipping a real product is completely different from running a notebook

I am a Mechanical Automation & Maintenance Engineer now specialising in Industrial AI. Most software people build AI apps. I build AI apps that understand real physical systems and real operational problems.

This is what 3 days of focused building looks like. 👇

🔗 Try it live: https://lnkd.in/dAFBcYM5

💻 GitHub: https://lnkd.in/d9cGAUcp

Mohammad Arshad

Bayut.com dubizzle Property Finder Dubai Land Department Emaar DAMAC Properties Better Home Group