PropTech.Brain - Contact Center AI Transformation

Project Name: PropTech.Brain - Contact Center AI Transformation Platform

The Challenge (Problem):

A large-scale property management contact center with 20+ operators was struggling with extreme operational complexity. Operators were forced to manage 15+ open tabs (Excel, Word, legacy databases) simultaneously to find information. This led to high error rates, slow response times, and a massive overhead of 5 people per shift just to manage CRM data entry and Telegram requests.

The Solution:

I architected and developed "PropTech.Brain" - a centralized AI-driven ecosystem that unified all fragmented data into a single "intelligent brain." The system features an autonomous execution layer that processes incoming requests, analyzes context, and takes action without human intervention.

Key Technical Features:

Dynamic Solution Chains (Personalized Scripts):

Developed a unique logic for individualized conversation scripts. The AI generates a tailor-made response strategy for each specific client based on their history, current building status, and the nature of the issue.

Autonomous CRM & TG Orchestration:

The system fully replaced human operators on CRM+Telegram shifts. It automatically creates, updates, or rejects requests based on real-time data analysis and routes them to the appropriate technical or maintenance teams.

Single-Window Knowledge Centralization:

Built a high-speed RAG (Retrieval-Augmented Generation) pipeline that indexed all 15+ fragmented data sources into a unified vector database (Qdrant), providing operators with near-instant access to any information.

Automatic Post-Call Documentation:

Implemented a voice-to-structured-data engine. After a call ends, the AI automatically extracts all relevant details (address, problem type, urgency) and populates the CRM fields, eliminating manual entry.

Proven Impact:

Workforce Optimization: Reduced the required headcount for CRM/TG shifts by 100% (5 out of 5 people replaced).

Complexity Reduction: Eliminated the need for 15+ external tabs, centralizing everything into one automated interface.

Operational Speed: Automated the full order lifecycle from initial contact to technician assignment.

Accuracy: Semantic duplicate detection prevented redundant task creation and optimized technician workflows.

The Stack:

Python 3.12, FastAPI, PostgreSQL (pgvector), Redis, SQLAlchemy, Mistral Large, Groq, Docker, Next.js 14.