A petroleum-engineering language-model family trained for technical reasoning and retrieval workflows.
Open research
Concept diagram. Instruction corpus → Fine-tuned Qwen3 → Retrieval-grounded answer.
Domain-adapted Qwen3 8B and 32B models for oil-and-gas knowledge work, coupled with a retrieval workflow for answers grounded in source material.
The problem
Generic models can miss technical vocabulary, procedures and context in industrial domains, yet confident answers are not enough for engineering decisions.
The approach
Construct a petroleum-engineering instruction corpus, fine-tune model variants, compare performance and connect the model to retrieval for source-backed responses.
Scope and limits
Use as research and engineer-reviewed decision support only; not for autonomous safety-critical or operational decisions.
