Privestate

Privestate is a privacy-preserving AI platform that enables attested users to access real estate insights using secure compute-to-data (C2D) jobs on Ocean Protocol Nodes. It empowers institutional investors, analysts, and developers to extract intelligence from sensitive property datasets—without ever exposing raw data or underlying algorithms.

How it works: 1. Dataset Publishing – A curated dataset of Dubai real estate listings (data.json) and a basic algorithm (average-price.py) are published via Ocean Node running on Fluence infrastructure. 2. Attestation Gate – Users must present a verifiable Self Protocol credential (e.g., investor status, region) before they can access or compute on the data. 3. Compute-to-Data Job Execution – Once approved, the user triggers a C2D job via Ocean Node’s free compute layer. The algorithm processes property data to return insights like average price per zone. 4. AI Post-processing – The result is forwarded to ASI1-mini for enrichment: LLMs summarize key trends, flag anomalies, or generate predictive commentary. 5. Result Display – A lightweight frontend (Next.js + Tailwind) shows the full user flow: wallet connect → attestation → compute job → enriched AI result.

Privacy by Design • No raw data or model is shared with the user or third parties. • All access is protected by decentralized identity credentials via Self Protocol. • Ocean Node ensures the dataset is only used inside sandboxed compute jobs.

Real-World Application For the demo, Privestate predicts average property prices across Dubai neighborhoods. In future iterations, it can be expanded with: • Advanced ML models (e.g., XGBoost) • Private, paywalled datasets • AI ranking/recommendation agents (e.g., “Find best ROI zone”)

Privestate demonstrates how decentralized infrastructure, AI, and privacy-first design can transform data access in real estate—without sacrificing compliance or control.

I built Privestate to explore what a decentralized, privacy-preserving AI platform could look like especially in the context of sensitive data like real estate.

The core of the system is powered by Ocean Protocol, specifically the free compute-to-data (C2D) feature of Ocean Nodes. I deployed an Ocean Node on Fluence, which let me run everything in a decentralized cloud environment with no centralized backend or server maintenance.

To control access, I integrated Self Protocol, allowing users to present verifiable attestations (like KYC or region). If the attestation is valid, they can submit compute jobs on the real estate dataset I prepared (data.json, containing actual Dubai property listings). If not, the system blocks them before they even reach the data.

For the AI side, I used a basic Python script (average-price.py) as the algorithm to run on Ocean. It computes average property prices by zone, price per square meter, and some simple aggregation stats. After the compute job runs, the output is sent to ASI1-mini, which adds an AI-powered summary layer—turning raw data into readable insights.

The frontend was built with Next.js, Wagmi, and Tailwind CSS. It’s intentionally minimal: just enough to show the flow from wallet connection, to attestation check, to compute job submission, and finally to viewing the enriched results. Everything runs directly in the browser and interacts with the Ocean Node and ASI1-mini via API.

The hackiest part was deploying the Ocean Node on Fluence—getting the node to behave correctly under compute constraints and still support C2D was a real challenge. But once it worked, it felt like a small victory toward decentralizing heavy AI infrastructure.

All in all, this project ties together decentralized compute, privacy-respecting access, and AI enrichment in a way that feels practical and forward-looking. And I did it all solo over a few intense days—definitely one of my favorite builds so far.