I automate the manual work slowing your team down.

End-to-end job management automation for an Australian field services firm. Handles job creation, technician scheduling, document generation, and financial sync across Outlook, NocoDB, SharePoint, and Xero, eliminating manual data entry across the entire ops cycle.

A field services business was managing jobs, technician scheduling, documents, and invoicing across disconnected tools. Data was being entered manually multiple times across systems, causing errors, delays, and missed billing.

A 90-node n8n workflow that connects every part of the ops cycle. Jobs flow from creation through scheduling, document generation, and financial sync automatically, with no manual data entry at any stage.

1. 01New job request triggers the workflow via Outlook email or manual entry
2. 02Job details are parsed and stored in NocoDB as the central data layer
3. 03Technician scheduling logic assigns the right resource based on availability and location
4. 04Job documents are generated automatically and stored in SharePoint
5. 05On job completion, financial data is synced to Xero for invoicing
6. 06Status updates flow back to the client automatically via Outlook
7. 07The full job lifecycle is tracked in NocoDB from creation to invoice

• →Manual data entry eliminated across the entire job lifecycle
• →Job creation to invoice handled automatically with no human touchpoints
• →Documents generated and stored without manual filing
• →Financial sync to Xero removes duplicate entry between ops and finance

• →90 nodes covering the complete field services ops cycle
• →NocoDB as a flexible no-code database at the centre of the architecture
• →SharePoint integration for automatic document storage and retrieval
• →Xero sync ensures finance team always has accurate job-based billing data

A government-adjacent digitization pipeline built to process approximately 3 million land survey documents. The system compresses, OCRs, and parses each document through a fully open-source stack, with preprocessing tuned specifically for real-world scan quality rather than generic OCR benchmarks.

A large volume of physical land survey records, roughly 3 million documents, existed only as scanned PDFs with no structured, searchable data behind them. Off-the-shelf OCR tools failed on the real scan quality: skewed pages, inconsistent lighting, and dense handwritten annotations that generic OCR benchmarks don't account for.

A fully open-source digitization pipeline that compresses, OCRs, and parses each document at scale. Vision-language models handle the documents generic OCR can't read, while a fine-tuned Tesseract instance handles high-volume standard text as the CPU workhorse, keeping cost manageable against a fixed project fee.

1. 01Documents are ingested and compressed for efficient processing and storage in Cloudflare R2
2. 02Preprocessing steps, deskew, denoise, CLAHE, adaptive threshold, are applied and tuned against real survey document scan quality
3. 03Each document is routed through the appropriate OCR engine: PaddleOCR-VL and Qwen2-VL-7B on GPU for complex or handwritten content, fine-tuned Tesseract on CPU for standard text
4. 04Extracted text is parsed into structured fields and written to PostgreSQL
5. 05Redis and Celery orchestrate the pipeline across compress, OCR, and parse stages so processing scales without manual intervention
6. 06The full pipeline runs end to end via a single chained process, confirmed working on real documents

• →Processes roughly 150 to 250 documents per active hour, peaking near 500, on the current setup
• →First successful end-to-end pipeline run confirmed on real survey documents
• →Preprocessing tuned specifically against real scan quality, not generic OCR test sets
• →Hybrid GPU/CPU OCR routing keeps processing cost sustainable at millions-of-documents scale
• →Fully open-source stack avoids per-document licensing costs that would be unworkable at this volume

• →Designed for ~3 million documents, a very different scale problem than a typical CRM or workflow build
• →Custom preprocessing tuned against real-world scan artifacts rather than clean benchmark data
• →Hybrid OCR strategy: vision-language models for hard cases, traditional OCR for volume, balancing accuracy against cost
• →Built and evaluated against multiple OCR engines (Azure DI, PaddleOCR-VL, olmOCR, DeepSeek-OCR) before settling on the current stack