Sharp Sotheby's International Realty
Multi-app platform powering brokerage operations across Cyprus, Hungary, and Kazakhstan.
React TypeScript Vite Tailwind Supabase PostgreSQL Deno Databricks Delta Lake OpenAI Claude Gemini HumaticAI LangChain Qwen3 Cursor Lovable Node.js Python Apache PM2 GitHub Twilio Microsoft WhatsApp MCP Zod OData RESO OAuth Linux
Channels (WhatsApp, Email, Web, Microsoft 365, Ads)
│
▼
Matrix Apps ── built from a shared App Template (React + Vite + shadcn/ui)
│
├── CDL-Connected Apps ─── Pipeline, Contact Mgmt, Broker Dashboard
│ read/write shared RESO DD 2.0 data layer
│
├── Domain-Specific Apps ── HRMS, Financial Mgmt, ITSM
│ own Supabase databases, shared SSO auth
│
└── AI Services ── Zoe Assistant, AI Web Chat, Blog Generator
context-aware copilots for brokers, managers, and visitors
│
▼
Supabase (CDL + Auth) ◄──► Databricks (DWH + ETL) ◄──► MLS / Portals
24 apps in the ecosystem — 11 live, 7 in progress, 6 planned.
| Layer | Technology |
|---|---|
| AI Development | Cursor (Claude Opus) — AI-native IDE for platform engineering |
| App Builder | Lovable — AI-powered app builder, generates full-stack apps |
| Frontend | React 18, Vite, TypeScript, shadcn/ui, Tailwind CSS, TanStack Query, Radix UI |
| Auth & IAM | OAuth 2.0 + PKCE, custom JWT, HMAC-SHA256, 5-level RBAC (self → team → global → org_admin → system_admin) |
| Backend | Supabase Edge Functions (Deno/TypeScript), PostgreSQL 15, Row-Level Security |
| Realtime | Supabase Realtime (WebSocket) — live updates across all apps |
| Data Standard | RESO DD 2.0 — canonical data layer, OData 4.0 Web API |
| Data Pipeline | Databricks — Medallion ETL (Bronze → Silver → Gold), CDC every 15 min, Delta Lake |
| AI/ML | AI Brokerage Copilot, LangChain, RAG (retrieval-augmented generation), semantic search, lead scoring, recommendation engine, personalization |
| NLP & LLM | OpenAI GPT-4, Claude (Anthropic), Google Gemini, Qwen3 embeddings, AI agents for Next Best Action & document generation |
| Conversational AI | HumaticAI — autonomous WhatsApp replies, personality profiling, lead qualification |
| Channels | WhatsApp Business API (Twilio), Microsoft Graph (Exchange, Calendar, AD), SMTP/IMAP |
| MCP Servers | Model Context Protocol — AI tools for CRM access (Qobrix), RESO-aligned workflows, schema introspection |
| Integrations | Qobrix REST API, Dash/Anywhere.com API, SIR syndication, Google/Meta Ads API |
| Infra | Self-hosted Linux (Debian), Apache, PM2, Node.js, Python 3.13, automated deploy via GitHub webhooks |
| Observability | Structured logging, PM2 monitoring, Supabase Dashboard, Databricks job metrics |
| i18n | i18next — English and Russian |
We don't take ourselves too seriously. It's extremely difficult to compete with someone building for pleasure and fun.
Sharp Matrix is not a product — it's an AI-native digital ecosystem that elevates luxury real estate brokerage to a completely new level. While others debated chatbot UX, we shipped 24 interconnected apps, autonomous assistants, and a unified data platform that runs a brokerage across three countries. Everything here — every line of code, every AI model, every pipeline — exists to make Sharp Matrix better.
- Platform, not a feature — Sharp Matrix is 24 interconnected apps (CRM, HRMS, Financial Management, ITSM, Pipeline, Client Portal), a RESO-standard data layer, a Databricks ETL pipeline, AI assistants (Zoe, Web Chat, HumaticAI), lead scoring, property matching, semantic search, and a recommendation engine — all wired together, sharing context, learning from every interaction.
- From language to agency — the ingredients for autonomous systems existed for years. Sharp Matrix combined them into a living platform integrated into WhatsApp, email, Microsoft 365, and daily broker workflows — where AI is not a tool you open, but a colleague that works alongside you.
- One data truth, three markets — RESO DD 2.0 as the canonical layer. One field name everywhere, every app, every market. Cyprus, Hungary, Kazakhstan share one platform with regional configuration, not code forks.
- The broker works with the client. The system works with the process. — AI Copilot handles data, timing, matching, and forecasting. Brokers focus on what only humans do well: build trust, negotiate, and close deals.
- Creative problem-solving over programming — Sharp Matrix agents demonstrate world knowledge that was never explicitly coded. They figure things out, improvise, and solve problems we didn't anticipate.
- Vibe coding > structured development — while big labs focused on language models, we built Sharp Matrix through iterative "vibe coding" — shipping fast, learning faster, and allowing more experimentation than any corporate environment would permit.
| Project | Description |
|---|---|
| Qobrix CRM MCP | Read-only MCP server for Qobrix CRM — 42 tools across 13 entity groups, aligned with RESO DD 2.0 canonical workflows. 148 tests. |
| MLS 2.0 Pipeline | Databricks ETL pipeline + RESO Web API for MLS data ingestion (Bronze/Silver/Gold). |
| Matrix Platform KB | Full knowledge base for the Sharp Matrix platform — architecture, data models, business processes, RESO DD mapping. |
| GitHub Watcher | Zero-dependency webhook server that auto-deploys repos on git push. No CI needed. |
- 90%+ process automation (from 20-30% today)
- Lead processing in 30 min (down from 4 hours)
- Sales cycle cut to 50-60 days (from 90-150)
- 3x website traffic, 3x marketing CTR, +160% deal close rate
- 7-phase rollout: Foundation → CRM Migration → AI Copilot → Client Portal → Marketing Automation → Analytics → Optimization