Client retained at $12K/month · Expanding to 3 additional warehouses · Phase 3: Computer vision quality inspection at receiving docks (in progress)
Who we worked with
A national logistics and distribution company operating 12 warehouse facilities across the United States, managing inventory for 200+ B2B clients with approximately 600 employees. At peak they processed 15,000+ orders daily across all locations.
Their 13-year-old WMS — built on PHP 5.4 with a MySQL backend and a desktop-only interface — had scaled from serving 3 warehouses to 12 without any architectural evolution, and was buckling badly under the weight of that growth.
A 13-year-old WMS becoming an operational bottleneck
Blind inventory, manual order intake, error-prone fulfilment, and inefficient picking were compounding daily — draining millions in avoidable costs.
Zero Real-Time Visibility
Managers had to call each facility individually to check stock. The system ran end-of-day batch updates only — no live picture of what was happening across 12 locations.
End-of-day batch only91% Fulfilment Accuracy
9% of orders shipped with wrong items, wrong quantities, or wrong addresses. The resulting returns and re-shipments cost $1.2M annually — and damaged client relationships.
$1.2M/yr error costSpreadsheet Forecasting
The VP of Operations ran quarterly inventory forecasts manually in spreadsheets. The result: $600K in excess safety stock sitting idle alongside regular stockouts on fast-moving SKUs.
$600K excess inventoryNo ERP Integration — All Manual
The PHP system had no API layer. Every client order arrived by email and required manual data entry, adding 2–3 hours of lag before fulfilment could even begin.
2–3 hr order lagInefficient Pick Routes
Picking routes were assigned randomly with no optimization. Warehouse pickers were walking an estimated 40% more distance than necessary — wasting labour hours on every single shift.
40% excess walkingNo Path to AI or Scaling
The CTO wanted AI-powered forecasting and pick-path optimization, but the legacy monolith had no API layer, no data pipeline, and ran on a single dedicated server with no room to grow.
Single server ceilingTwo phases, one intelligent warehouse platform
We ran modernization and AI integration in overlapping phases — rebuilding the WMS foundation while layering demand forecasting and route intelligence on top as the new platform came live.
- Cloud-native WMS on AWS ECS Fargate — rebuilt the PHP monolith as Node.js microservices, replacing all 12 fragmented on-premise installations with a single elastic architecture
- Real-time inventory tracking — every scan (receive, pick, pack, ship) updates the central database instantly and propagates to all dashboards within 3 seconds
- React web app + React Native mobile app for warehouse floor workers — barcode scanning, pick confirmation, and exception handling all on handheld devices
- RESTful API gateway enabling direct ERP integrations with clients — automated order intake replaces all email and manual data entry workflows
- Cross-warehouse visibility dashboards — managers see real-time inventory levels, order status, and workforce utilisation across all 12 facilities from one screen
- CI/CD pipeline with automated testing — production deployments in under 15 minutes with zero downtime across the entire network
- Demand forecasting engine (Python + Prophet) trained on 4 years of per-client historical order data — incorporates seasonal patterns, promotional calendars, and macroeconomic indicators
- Weekly automated replenishment recommendations per SKU per warehouse, delivered to the VP of Operations every Monday morning with variance explanations
- Intelligent pick-path optimization — modified travelling salesman heuristic calculates the shortest walking route through the warehouse for every batch of orders, before pickers leave the station
- AI anomaly detection — flags unusual patterns in real time: unexpected demand spikes, inventory shrinkage trends, and receiving discrepancies, alerting managers before problems escalate
- Predictive labour planning — based on forecasted order volume, recommends shift staffing levels per warehouse 2 weeks in advance, reducing over- and under-staffing variance
Full-stack cloud-native logistics architecture
Results that changed how they operate
Measurable improvements across accuracy, productivity, inventory costs, and client onboarding — all delivered within 22 weeks with zero disruption to daily fulfilment operations.
Accuracy
Increase
Stock Eliminated
Same Budget
6 engineers, one warehouse-to-cloud delivery
A lean, high-output team covering cloud architecture, full-stack development, mobile, machine learning, and DevOps — structured to roll out across 12 live warehouse facilities without missing a single shipment.
Solution Architect
AWS cloud architecture, microservices design, API gateway strategy, phased rollout planning
Senior Node.js Developer × 2
Microservices backend, real-time inventory engine, ERP integration API layer, SQS event processing
React / React Native Developer
Web management dashboards, mobile barcode scanning app, pick confirmation UI for floor workers
ML Engineer
Python, Prophet demand forecasting, pick-path optimization algorithm, scikit-learn anomaly detection
DevOps / QA Engineer
AWS ECS Fargate, Terraform, GitHub Actions CI/CD, automated testing, 12-facility rollout coordination
22-Week Full Delivery
Pilot at 2 warehouses · Phased rollout to all 12 locations · Zero disruption to daily order fulfilment throughout
What the client said
"i-verve replaced a system we'd been fighting for over a decade — and did it without losing a single day of operations across 12 warehouses. Our error rate dropped from 9% to under 1%, our pickers are 38% more productive, and we've freed up $380K in inventory capital we were just sitting on. The AI forecasting alone has transformed how we plan procurement. This wasn't an IT project — it was a fundamental change in how we run our business."