labs@atheron:~/industries/mining/ai$ train agent --grounded

AI for geologists

Decades of exploration data, searchable and cited.

Exploration companies inherit decades of data from past owners, much of it scanned. An assistant that reads all of it and cites the page saves geologists from rebuilding history by hand. Because unpublished results are material information, we usually run it on an open-source model you control.

Start a project See the example projects

// what matters here

What is different in this industry.

Unpublished data stays private

The model runs in your own cloud account, so results that have not been disclosed never go to an outside provider.

Every fact has a page

Each answer cites the report, page and table, so a geologist can check it before relying on it.

Scans included

Old reports are read with OCR and layout models, tables and logs included, and the scan is kept beside the extracted text.

Checked by a geologist

Extracted intercepts land in tables for a geologist to review, never straight into the resource model.

// example projects

Priced examples for this service.

Each one is hypothetical, labelled as such, and priced live from our rate card. Open any of them in the estimator and make it yours.

Example projectWeb appAIOpen-source model

Geology assistant for old reports

Hypothetical. Not a client, not a result.

// The problem

An exploration company has acquired properties with decades of drill logs, assays and technical reports from past owners, mostly scanned, and its geologists spend days rebuilding what is already known about a target.

// What we would build

An assistant that reads the scanned reports, logs and assay certificates, answers questions about a property with the page each fact came from, and pulls drill intercepts into tables for geologists to check. It runs on an open-source model in your own cloud account, because unpublished results are material information that should stay under your control.

// What is in it

Scanned reports, logs and assay certificates read and indexed, tables included
Answers that cite the report, page and table
Drill intercepts extracted into tables for a geologist to check
Questions across every property in the portfolio
Evaluation set of questions with known answers, run before every change
An open-source model on GPUs in your own cloud account

// Stack

Qwen or Llama (open-source)
vLLM
OCR with layout models
PostgreSQL with pgvector
Next.js
AWS

// estimate

Build: ≈ US$92,100 to US$141,000, delivered within 22 weeksCAD 131,100 to 200,500
Hosting: ≈ US$6,710 a monthCAD 9,550 a month
Support: ≈ US$2,500 a monthCAD 3,565 a month

Prices in your currency are estimates from today's Bank of Canada rate. All invoicing is in CAD or USD.

AI route: Open-source model
Model running cost: ≈ US$6,530 a monthCAD 9,300 a month

Timeline by milestone

Discovery: 2.9 to 3.4 weeks
Specification and evaluation plan: 0.4 weeks
Design approved: 2.4 to 3.4 weeks
Core features: 1.4 to 1.9 weeks
Full build: 0.9 to 1.4 weeks
Working pilot: 3.9 to 6.9 weeks
Testing and fixes: 0.9 to 1.9 weeks
Launch: 0.9 to 1.4 weeks
Production: 0.9 to 1.4 weeks

How it is paid

Deposit 20%: CAD 26,160 to 40,040
Discovery 1.8%: CAD 2,354.40 to 3,603.60
Specification and evaluation plan 10.5%: CAD 13,734 to 21,021
Design approved 1.8%: CAD 2,354.40 to 3,603.60
Core features 5.6%: CAD 7,324.80 to 11,211.20
Full build 3.7%: CAD 4,839.60 to 7,407.40
Working pilot 31.7%: CAD 41,463.60 to 63,463.40
Testing and fixes 1.8%: CAD 2,354.40 to 3,603.60
Launch 1.8%: CAD 2,354.40 to 3,603.60
Production 11.3%: CAD 14,780.40 to 22,622.60
Holdback, 30 days after launch (10%): CAD 13,080 to 20,020
GPU time for training and testing, at cost: CAD 300. Billed up front, at cost, outside the milestones

Start from this example

// questions

Questions we get about this.

Can the assistant write our NI 43-101 report?

No. Technical reports are prepared and signed by qualified persons. The assistant helps them find and check what is already on file.

How does it handle poor scans and handwriting?

Reasonably for typed reports, less well for handwriting. Pages it cannot read with confidence are flagged rather than guessed.

Why not a frontier model for the geology data?

It would be quicker to start, but every document and question would go to the provider. For unpublished results that is usually not acceptable, so we price the open-source route first.

Back to the industry: Mining and resources About the service: AI agents and models

// next

Not quite your project?

Tell us what you have in mind. We will come back to you with a range and the questions that would narrow it. Or book a call and talk it through.

Start a project Book a call