Enterprise AI & SaaS
Production

Mahdar AI - Meeting Intelligence Platform

Enterprise meeting intelligence platform with real-time transcription, RAG-based search, and AI-driven insights, deployed on-premises for government and enterprise clients in Saudi Arabia.

LOOP
2024 - 2025
Role:Backend Engineer
Project Description

Contributed to Mahdar AI, an enterprise-grade meeting intelligence platform designed for secure, on-premises deployment in government and high-security enterprise environments. The platform provides real-time, speaker-aware transcription across multiple meeting platforms, combined with powerful RAG-based search capabilities enabling users to query past meetings and retrieve context-aware answers. Built with security, scalability, and data sovereignty as core requirements, the system runs entirely air-gapped in client infrastructure including Saudi government Ministry-level deployments.

Architecture

Pattern:

Kubernetes-native microservices with event-driven processing

Data Flow:

Meeting Audio → Transcription → Speaker ID → Storage → Indexing (Text + Vector) → Search/Analysis
Key Features
  • Real-time speaker-aware transcription
  • Multi-platform meeting integration
  • RAG-based semantic search over meetings
  • Context-aware query answering
  • Automated meeting summaries
  • Action item extraction
  • Smart tagging and categorization
  • Multi-workspace architecture
Impact & Achievements

Impact:

  • Enabled government entities to leverage AI meeting intelligence with data sovereignty
  • Deployed at Saudi Ministry-level environments with highest security clearance
  • Processed 5000+ daily meetings with real-time transcription
  • Provided searchable knowledge base from organizational meetings
  • Reduced meeting follow-up time by 50% through automated summaries

Key Achievements:

  • Successfully deployed in air-gapped government environments
  • Achieved sub-200ms real-time transcription latency
  • Built scalable RAG search over large-scale meeting data
  • Designed multi-workspace architecture for complete data isolation
  • Contributed to Ministry-level government deployments in Saudi Arabia
Technologies Used
Python
FastAPI
PostgreSQL
pgvector
Redis
Elasticsearch
Kubernetes
Docker
RabbitMQ
Whisper (Speech-to-Text)
Local LLMs
WebRTC
Prometheus
Grafana

Interested in Similar Projects?

I'd love to discuss how I can bring similar expertise and architectural thinking to your projects.