Documentation
BoreholeAI reads borehole logs — scanned, digital, handwritten, or photographed — and extracts geotechnical data into structured formats like Excel and AGS. Use the web app for interactive extraction, or the Python SDK and REST API to automate at scale.
What Gets Extracted
BoreholeAI processes each page of your borehole log through a multi-stage pipeline that combines OCR, computer vision, and spatial reasoning. The following data is extracted automatically.
Ground Profile
Layer depths, thicknesses, material types, geological descriptions, consistency and density classifications, and rock strength — parsed from complex log formats into clean, structured tables. Multi-page logs are handled automatically, with depth continuity maintained across pages.
In-Situ & Laboratory Tests
SPT N-values, RQD, TCR, SCR, UCS, Is50, Cu (undrained shear strength), Atterberg limits (LL/PL/PI), and Emerson class are extracted with depth correlations. Results are grouped by test type for easy analysis.
Groundwater
Groundwater strike levels, rest levels, and monitoring data are captured with dates and depths. Multiple observations per borehole are supported.
Metadata
Borehole ID, project name, location coordinates, drilling dates, contractor details, drilling method, inclination, and other header information — extracted from the first page of each log.
Output Formats
Every processed borehole log produces four output files, available for download from the web app or returned by the API.
| File | Format | Contents |
|---|---|---|
| Borehole_ground_profile.xlsx | Excel | Structured ground profile with material descriptions, depths, geology, consistency/density |
| Borehole_test_data.xlsx | Excel | All test results — SPT, UCS, Is50, Cu, groundwater — in tabular format |
| Borehole_ags4.ags | AGS4 | Industry-standard AGS4 data transfer file with 11 groups (PROJ, TRAN, LOCA, GEOL, SAMP, ISPT, RUCS, RPLT, IVAN, IPEN, WSTG) |
| *_annotated.pdf | Original document with depth boundaries and material labels overlaid — a full audit trail |
How It Works
BoreholeAI uses a multi-agent, multi-stage agentic system that combines engineering deterministic algorithms, customised deep learning models, and computer vision with AI-assisted document understanding.
Your documents are never read directly by AI models. We apply an OCR intermediary layer so that AI only works with extracted text and layout information, never with your original files. The structured output is driven by spatial reasoning and rule-based logic, not generative models.
The system handles complex layouts, varying scales, multi-page logs, and inconsistent formatting. We have tested extensively across a wide range of borehole log formats and consistently achieve 95–100% accuracy.
Supported Input Formats
Upload any borehole log as a PDF or image. Any template, from any company, simple or complex. Single-page and multi-page logs are both supported. Available formats and upload limits vary by interface.
| Interface | Accepted files | Per-file limit | Other limits |
|---|---|---|---|
| Web app (Dashboard) | PDF, PNG, JPG/JPEG, WebP | 200 MB | Multiple files per upload |
| Playground | PDF, PNG, JPG/JPEG, WebP | 3 MB | Maximum 4 pages; 2 uploads per 24 hours |
| Python SDK / REST API | PDF, PNG, JPG/JPEG, TIFF, WebP | 90 MB | — |
Current Scope & Limitations
- Standard alignment: Not limited to Australian/NZ logs. Validated most extensively against Australian Standard AS 1726 formats; logs following other national standards are supported. If your logs use a non-standard or customised layout, contact us.
- Document types: Borehole logs (BH), pavement core logs (PCP), and test pit logs (TP). Other geotechnical documents (e.g. CPT plots, lab reports) are not currently supported.
- Languages: Not limited to English — logs in other languages are supported.
- Units: Metric only — depths and dimensions must be in metres.
- Scans and photos: Scanned copies and photos taken directly (e.g. a phone camera shot of a printed or handwritten log) are both supported. Keep the page flat and roughly square-on; heavily tilted or skewed captures reduce accuracy.
- Watermarks: Heavy watermarks, especially tilted ones, will reduce accuracy. Upload a clean copy where possible.
- Handwriting: Handwritten logs are supported. Extremely unclear handwriting may still affect results.
Next Steps
Ready to get started? Follow the quickstart guide, or jump straight to the web app or Python SDK.