Bulky
Enterprise Mining Data Platform
Replaced Excel macro workflows across 6 mining sites with a browser-based platform that validates and uploads operational data in real-time. Started as a single-site upload tool, then architected it as a multi-site platform with async SQS processing, concurrent job execution, and automatic delivery to on-premise systems.
An ASX-listed mining company relied on Excel files with VBA macros to submit operational data to their mining management system. The macros caused versioning issues, security restrictions under Office 365, and inconsistency between sites.
Sole engineer, 92% of 593 commits. Architecture, infrastructure, CI/CD, and deployment across DEV/STG/PRD. Platform successful enough that a dedicated engineer was hired to take it over.
Highlights
In-browser spreadsheets replacing Excel
Browser-based spreadsheet engine replaces Excel macros entirely. 8 data modules (DigBlock, ExPit, Crushing, StockYard, Rehandle, Samples, Haulage, Transactions) with validation rules baked into each template. Errors caught before anything touches the downstream system
Async processing with real-time status
SQS-based job queue with concurrent processing (up to 3 files simultaneously). Multi-stage validation: schema check, business rule validation, then CSV transformation. WebSocket pushes live status updates. ALB in private subnets with CloudFront VPC Origins
Performance engineering
13 performance commits in 2 days: Redis caching across 12+ endpoints, row virtualisation for large datasets, lazy-loaded modals, WebSocket reconnect with jitter backoff. Server-side pagination with DynamoDB GSI-based querying and prefetched next page
On-premise delivery with recovery
Processed CSVs delivered to on-premise watch paths via SMB/CIFS. Automatic retry on socket errors, startup recovery for stuck files, and event-driven log monitoring for delivery confirmation. ARM64 ECS tasks for cost savings