Meana Petro
KC-03FronteraOla 3IA

Asistente Multimodal de Subsuelo

En una frase

Understands log images, maps and cross-sections alongside text, so you can ask a question about a figure rather than only about a paragraph.

El problema

A large share of subsurface knowledge lives in figures: log displays, structure maps, seismic sections, core photographs. Text-only retrieval is blind to all of it, and the figure caption is rarely enough.

Cómo funciona

Vision-language models index figures alongside text, with domain-specific handling for log tracks, map annotations and section displays. Answers reference the specific figure and region, and the system is deliberately conservative about reading values off a plotted curve.

Datos necesarios

  • Figure-rich document archive with adequate scan resolution
  • Annotated figure examples for domain tuning
  • On-premise GPU capacity for vision model inference
  • Expert evaluation set covering figure-based questions

Qué obtiene

  • Answers referencing specific figures with the relevant region highlighted
  • Figure search across the archive by content rather than caption
  • Cross-references between figures and related text sections

Qué no puede hacer

Reading precise numeric values off a plotted curve is unreliable and we do not present it as a feature — the model locates and describes figures rather than digitising them. Scan quality is the binding constraint: a low-resolution scan of a log display is unreadable to the model for the same reason it is unreadable to a person.

Cada página de capacidad declara una limitación real. Si un proveedor no puede decirle dónde termina su producto, no le están diciendo la verdad sobre dónde empieza.

Capacidades relacionadas

  • KC-01Fundacional

    Asistente de Conocimiento Técnico (RAG)

    Ask a question about your own field in Persian, Arabic, English or Spanish and get an answer with the report and page it came from.

    Conocimiento y CopilotoIA
  • SR-04Intermedia

    Análisis de Roca Digital e Imágenes de Núcleo

    Measures porosity, grain size and fracture density from core photographs and CT scans instead of from a description written twenty years ago.

    Inteligencia de Subsuelo y YacimientoIA
  • KC-02Frontera

    Modelo de Lenguaje Específico del Dominio

    A model further trained on petroleum engineering text, so it uses the field’s vocabulary correctly in Persian and Arabic as well as English.

    Conocimiento y CopilotoIA