Client Results

The Situation

A B2B analytics platform serving mid-market companies was losing enterprise prospects during sales demos. Their main dashboard - the first thing potential customers saw - took 8-12 seconds to load with production-scale data. The engineering team had tried adding indexes and caching individual queries, but the core problem was architectural.

The platform ran on Rails 7 with PostgreSQL, processing 15M+ events daily across 200+ tenant accounts. Monthly AWS spend had grown to $14,000 and was climbing with each new customer. The CTO needed the dashboard under 3 seconds and infrastructure costs under control before their Series A conversations.

What I Did

• Ran a full performance audit: profiled every dashboard query, mapped N+1 patterns, and identified 3 queries responsible for 80% of load time
• Restructured the heaviest aggregation queries using materialized views and partial indexes, eliminating full-table scans on a 400M-row events table
• Implemented multi-layer caching with Solid Cache - Russian doll caching for dashboard widgets, with background refresh so users never see stale data
• Replaced the oversized RDS instance with read replicas and connection pooling via PgBouncer, right-sizing the primary database
• Set up query performance monitoring with pg_stat_statements dashboards so the team could catch regressions early

The Situation

A growing online marketplace connecting local artisans with international buyers was running Rails 6.1 on Heroku with Sidekiq Pro for background jobs. Their monthly infrastructure bill had crossed $8,200 - split between Heroku dynos ($3,400), Redis for Sidekiq and caching ($2,100), and various add-ons ($2,700). Order processing during peak hours was hitting Redis memory limits, causing failed jobs and delayed shipment notifications.

The two-person engineering team was spending 15+ hours per month managing Sidekiq queues, restarting failed jobs, and dealing with Redis connection issues. They needed to reduce costs and operational overhead so they could focus on product features instead of infrastructure firefighting.

What I Did

• Upgraded the application from Rails 6.1 to Rails 8 incrementally - 6.1 to 7.0, 7.0 to 7.2, then 7.2 to 8.0. Fixed deprecations and tested each step against their full test suite
• Migrated 47 Sidekiq jobs to Solid Queue with zero downtime - ran both systems in parallel for 2 weeks to verify job processing parity before cutting over
• Replaced Redis caching with Solid Cache, eliminating the dedicated Redis instance entirely. Configured cache expiry and size limits appropriate for their traffic patterns
• Moved from Heroku to a VPS deployed with Kamal 2 - configured automated deploys, health checks, SSL, and zero-downtime rolling restarts
• Set up Mission Control Jobs dashboard for the team to monitor queues, plus alerting for failed jobs via Slack integration

The Situation

A logistics coordination company managing freight forwarding across the UAE and GCC region was running their entire operation on shared Google Sheets and WhatsApp groups. With 40+ active shipments at any time, they were losing track of status updates, missing customs deadlines, and spending hours each day on manual data entry. They had been quoted $180,000+ by two agencies for a custom platform with a 6-month timeline.

The founder needed a working system - not a prototype - within 3 months, with a budget under $50,000. The platform had to support real-time status tracking, document management, multi-role access for their team and external partners, and automated notifications for deadline-sensitive milestones.

What I Did

• Built the platform on Rails 8 with Hotwire (Turbo + Stimulus) - no React, no Vue, no separate frontend build step. Real-time updates via Turbo Streams over Action Cable for live shipment tracking
• Designed a role-based access system with Rails 8 authentication - internal team, partner companies, and read-only client portals, each with appropriate permissions
• Implemented document management with Active Storage, automated deadline tracking with Solid Queue background jobs, and email/SMS notifications for critical milestones
• Deployed on a single VPS with Kamal 2 - total hosting cost under $80/month including the database, file storage, and background jobs
• Built a responsive interface with Tailwind CSS that works on desktop for office staff and mobile for warehouse and port teams