Loading...
Works

The Flex PMS Sep 2025-Present

Enterprise-grade Property Management System serving 500+ properties with 15K+ daily transactions across London, Paris, and Algiers. Built with micro-frontend architecture, real-time multi-tenant isolation, and production-grade observability stack that reduced MTTD from 45min to 8min and MTTR from 2hr to 25min.

  • PlatformEnterprise SaaS Web Application
  • StackReact, FastAPI, PostgreSQL, Redis, Docker, Kubernetes
  • Scale500+ Properties, 15K+ Daily Transactions, 3M+ Monthly Transactions
  • SRE MetricsP99 400ms, 99.7% Message Delivery, MTTD 8min, MTTR 25min
The Flex PMS Dashboard

SRE & Infrastructure

  • Tenant Data Isolation: Designed PostgreSQL Row-Level Security policies with PgBouncer connection pooling, eliminating cross-tenant data leaks and reducing auth overhead by 70%
  • Performance Optimization: Diagnosed P95 latency spikes (800ms+) caused by N+1 queries; implemented Redis write-through cache with 15-min TTL and batch prefetching, dropping P95 to 85ms and cutting database load by 45%
  • Observability Stack: Built structured JSON logging → Loki → Grafana dashboards with 12 SLI panels tracking error rates, latency percentiles, and saturation against defined SLOs
  • WebSocket Reliability: Debugged connection drops under load (TCP keepalive misconfiguration); redesigned with heartbeat mechanism, exponential backoff reconnection, and Supabase Realtime channels—achieved 99.7% message delivery
  • CI/CD Pipeline: Led toil reduction initiative; built GitHub Actions pipeline with parallel test execution, Docker layer caching, and blue-green deployment—now ship 3x daily with zero rollbacks
  • Micro-Frontend Architecture: Split monolith into 5 independently deployable modules behind Nginx ingress controller with path-based routing—enabled parallel team development, reduced deploy conflicts by 90%
  • Database Optimization: Optimized schema design with composite indexes on (tenant_id, created_at), query result pagination, and Redis cache warming on login—dropped P99 load time to 400ms for 3M+ monthly transactions
  • Security Implementation: Implemented defense-in-depth: OAuth 2.0 PKCE flow, JWT with 15-min expiry + refresh rotation, RBAC with 8 granular permissions, and immutable audit log (append-only table)—passed SOC 2 Type I readiness

Technology Stack

  • Frontend: React, TypeScript, Vite, TanStack Query, Tailwind CSS, Chart.js
  • Backend: FastAPI (Python), SQLAlchemy, Pydantic, async/await architecture
  • Database: PostgreSQL with Row-Level Security, PgBouncer connection pooling, composite indexes
  • Caching: Redis write-through cache with TTL, batch prefetching, cache warming strategies
  • Infrastructure: Docker, Kubernetes, Nginx ingress controller, GitHub Actions CI/CD
  • Observability: Prometheus, Grafana, Loki, OpenTelemetry, structured JSON logging

Key Metrics

  • Latency: P95 dropped from 800ms to 85ms (10x improvement)
  • MTTD: Reduced from 45min to 8min (82% improvement)
  • MTTR: Reduced from 2hr to 25min (79% improvement)
  • Auth Overhead: Reduced by 70% with RLS + connection pooling
  • Database Load: Reduced by 45% through caching strategies
  • Deploy Frequency: 3x daily with zero rollbacks
© 2026 Tasnimul Mohammad Fahim. All Rights Reserved.