Geoscience · AI mineral exploration

MinersAI

Data science R&D, machine learning and backend and ETL platform development since 2024, bridging the client's software engineering and geoscience teams.

The challenge

MinersAI gives mineral exploration companies a platform with geological data and AI tools. It needed data science and backend work in a complex environment that spans software engineering and geoscience.

What we did

  • Started with a discovery phase in 2024, focused on the quality and availability of geological data on the platform.
  • Built the backend, machine learning and ETL layer alongside the client's engineers. It now runs across three AWS environments and takes single uploads over 1 GB.
  • Rewrote the chunking step so the largest dataset so far, 1.3 GB of geochemistry with 561,232 features, no longer fails at AWS Lambda's 15-minute limit. It parses in 2.84 minutes, down from 4.35, with byte-identical output across all 281 chunks.
  • Stored drillhole assay tables as Parquet, about 6.9 times smaller than the same 14,786 rows in CSV.
  • Load-tested the platform at around 50 to 60 concurrent requests per Lambda, and sized an on-premises deployment for a client in Saudi Arabia at 2 to 4 GPUs, against a vendor quote of 18 H200s.
  • For a national geological survey bid, processed 27,591 files (99.9 GB), including 638 PDFs with 35,556 pages, cut 802 documents to about 619 by removing duplicates, and extracted 517 drillholes, delivering 203 of them with 3,734 intervals across 8 grid systems.
  • Delivered work the internal team could keep building on.

Metric agreed

The client's team can extend the work without Huitzi

The result

MinersAI's internal team consistently builds and iterates on what Huitzi delivers. The partnership continues.

“Their work is high-quality and reliable, and, just as importantly, maintainable. Our internal team has consistently built on and iterated on what they deliver without friction.”

Mason Dykstra, CEO, MinersAI

Technologies

  • Data science
  • Machine learning
  • Backend development
  • ETL
  • AWS Lambda
  • Parquet

Related services