Toptal reduces false positives by 93% for AI-powered mineral rights title acquisition platform.
An oil and gas technology company partnered with Toptal to improve the accuracy and reliability of an LLM-based platform designed to automate mineral title determination across hundreds of thousands of land records.
Client
A US-based oil and gas technology company offering AI-powered mineral title determination for land records.
Industry
EnergyDelivered Services
Get a free consultation now
Schedule a CallChallenge
Mineral title determination required analysts to manually review decades of scanned deeds and reconstruct ownership chains, making the process slow, expensive, and prone to error. Automating the process with LLMs introduced another challenge: reliably validating output across more than 200,000 documents.
Solution
Ground-truth AI Evaluation
Toptal built a queryable database of 2,000+ attorney-validated answers and a regression framework that measured every pipeline update against known results, exposing errors before they reached production.
Automated Title Resolution
Toptal helped translate complex legal doctrine into structured, testable rules, using SQL-based analysis and knowledge graph concepts to resolve ownership, succession, and identity across complex title histories.
Outcome
Improved Accuracy and Reliability
Toptal helped reduce the false-positive rate from approximately 27% to 2% while maintaining roughly 93% recall. Conveyance-party capture also improved from approximately 85% to 96%.
Increased Processing Confidence
The evaluation framework supports version-tracked runs across more than 200,000 documents, catching regressions before release and establishing a reliable foundation for expanding the title platform into new markets.
Get a free consultation now
Schedule a Call