AI ENGINEER
The Converge AI Engineer
The AI Engineer is the part of Converge that reads your lab report, builds the fluid with the same equation-of-state engine you use by hand, runs the calculations you ask for and explains the results. You decide what gets published.
FOUR STEPS
How does the AI Engineer build a fluid model?
It works in four steps: it reads the lab report, builds the fluid with the Converge equation-of-state engine, runs the calculations you ask for and hands you a draft to review. The steps run in order, and you can stop after any one of them and carry on by hand.
- 01 · PDF · XLSX · DOCX
Read the report
Upload the lab report, or paste a table. The AI Engineer pulls out the composition, plus-fraction properties, reservoir temperature and each experiment's measurements.
- 02 · EOS
Build the fluid
It sets up the equation of state, splits and lumps the plus fraction, and calculates saturation pressure and the lab experiments with the Converge engine.
- 03 · CALCULATIONS
Run what you ask
Ask for a phase envelope, a flash, a CCE or a separator test. It runs the calculation with the Converge engine in your project and explains the result next to the lab values.
- 04 · DRAFT
You review
The result is a draft model with every input it read from the report listed. Accept, change or discard it before anything is published. Tuning stays with you, in the live tuning screen.
SIGN-OFF
Rules it follows
An engineer has to sign off a PVT model. The AI Engineer is built to make that sign-off quick, not to skip it.
- It shows every value it read from the report, so you can check it against the source page.
- It says when the report is missing something the model needs, instead of filling the gap silently.
- It quotes results from engine calculations it runs in your project, and you see each calculation as it runs.
- Its work is a draft. Your published model changes only when you accept it.
PROVIDERS
Model providers
Use the provider your company already approves. If a provider is not configured, Converge says so; it never sends your request to a different one. On an on-premises install, the AI Engineer can use a model you host yourselves.
- Claude
- Gemini
- OpenAI
- Kimi