Mining Tech Leader Adds Computer Vision to Generative AI Research Platform

A Wealth of Mining Data Trapped in Paper Records

Prospector has already modernized the way mining companies, investors, and researchers work by making troves of public information easily accessible through its searchable database. In addition to public data, mining companies of all sizes have enormous volumes of proprietary data about their own holdings. Data from hundreds of years of legacy maps, drill core samples, ore volume calculations, and countless other details have the potential to inform better business decisions—if leaders could access the insights.

Unfortunately, many of these records sit isolated as paper documents or trapped in PDFs archived on private servers. Relevant information is often buried within documents that are, on average, 800 pages. Even the largest, most tech-savvy mining companies struggle to convert their documents into manageable data.

Prospector had a vision to make all mining data discoverable, regardless of the data type. To deliver on that vision, Prospector partnered with Striveworks to create OpenMine.AI—a platform that uses machine learning to marry publicly available data with companies’ own proprietary data, delivering actionable and easy-to-interpret intelligence.

Generative AI Streamlines Mining Analysis

First, Prospector combined its rich database with generative AI, letting customers ask questions in plain English. The chat interface, powered by a large language model (LLM), is capable of answering challenging questions such as, “What site in Chile has the richest copper reserves?” or “What price of nickel was used to determine how long this mine will last?”

Knowing that the key to unlocking value from customer archives was to extract data from images and return relevant text, Prospector then turned to the leader in machine learning operations (MLOps) and computer vision, Striveworks.

Computer Vision Extracts Insights Locked in Legacy Documents

Striveworks has a long track record of building machine learning workflows to solve customer challenges, and the company’s strengths with computer vision were well suited to the task of identifying and interpreting text from image files.

Striveworks fine-tuned and implemented the YOLO algorithm for this purpose. The enhanced model detects, extracts, and classifies images from mining imagery. Scanned maps, for example, now yield critical information, including site perimeter in latitude and longitude.

A variety of image types (including geophysical, photo, claims data) automatically detected and classified by Striveworks' computer vision models.

The object detection workflow improves analyst efficiency by automatically extracting images from dense technical documentation—a process that previously required sifting through multiple documents (some hundreds of pages long) to locate maps, geophysical images, drill hole diagrams, figures, and tables in order to confirm their relevance and interpret the text. OpenMine.AI now returns newly extracted text and images to users—alongside industry data from public regulatory filings—drastically simplifying search and filtering.

Expanding Prospector’s Success With the Fortune Global 500

Two Fortune Global 500 mining companies have engaged Prospector to pilot OpenMine.AI, with promising co-development opportunities to refine and expand the research platform to better suit their needs.

“Thanks to Striveworks’ data science capabilities, specifically model development and fine-tuning, I see the potential for OpenMine.AI to scale fast enough to serve the entire mining industry,” says Prospector Founder and CEO Emily King.

“By embedding Striveworks’ MLOps and data science power into the OpenMine.AI platform, Prospector can focus our resources on growing our business and serving our customers.”