The lab works on questions where a dashboard is not enough: what is happening, what can be measured, and what should happen next?
A prediction is only useful when the team understands its inputs, uncertainty, failure modes, and operational consequence.
Physics-informed Bayesian virtual sensing for wax deposition from routine oil-well and pipeline telemetry. The work treats wax state as a partially observed inverse problem with uncertainty and abstention.
Can an operator estimate a hidden flow-assurance restriction from routine measurements before intervention becomes expensive?
Next artefactValidated state-estimation tool and intervention decision path.
Digital-twin and production-intelligence work for petroleum engineering decisions, with a focus on how field behaviour becomes a usable engineering surface.
How can production and well signals become a decision surface instead of another isolated dashboard?
Next artefactOperator-facing model, evidence layer, and decision support.
A provenance-first knowledge system for raw evidence, canonical sources, synthesised notes, citations, and explicit contradictions.
How can an AI workflow remain useful when its claims must remain traceable?
Next artefactReliable evidence capture and maintenance workflow.
A language-agnostic AI study platform for learning, revision, tests, exam preparation, progress, and guardian clarity.
What does a study system look like when classroom help, personal notes, curriculum, exams, and guardian accountability share one product foundation?
Next artefactAI classroom and unified study journey.
Define the operational decision before selecting a model.
Check whether the available evidence can support the output.
Keep alternative explanations alive and abstain when needed.
Turn a model into a tool someone can use responsibly.
We can help turn it into a research plan, an observability audit, a baseline model, or a product that earns its place in the workflow.
Start a research brief