Compute for oil and gas apps, scale up in minutes
Reservoir simulation, history matching, decline-curve models and the machine learning on top of them, on machines PERD starts for you and shuts down when you are done, with a GPU when the model needs one. Each one is an app on a desktop in your browser: nothing to install, nothing lost when the connection drops.
Verify a phone number, then continue with Google.
- Workflows
- Decline curve · History match
- Machine
- GPU T4 · 4 vCPU / 16 GiB + 1× NVIDIA T4 (16 GB)
- Lease
- 3 days, renewed while the desktop is open
- Users
- 3, invited by email
Sized for the job
Each workflow declares the machine it needs, from a small CPU to an NVIDIA T4 GPU, and its workspace starts on it in minutes. Nobody sizes a server.
Nothing lost on a dropped connection
Close the laptop mid-run, or lose the link at site. Reconnect and every result is waiting, even for jobs that finished while you were away.
Nothing left running
Every workspace runs on a lease, so forgotten machines shut themselves down instead of piling onto your bill.
Three steps from a workflow to results
No cloud account, no server to set up, nothing to install.
Pick a workflow
Choose from the store, or from what your organization publishes for its own people. Each workflow carries its documentation, its versions and the machine it needs.
Open it as an app
Launch a workspace and each workflow is an icon on a desktop in your browser. Open one in a tab, upload a deck or a production history, press Run.
Watch it run, from anywhere
Losses, metrics and the reservoir grid stream to everyone watching, and are there again when you reconnect. When the lease ends, the machine shuts down.
For the people who run the models, and the people who build them
Engineers, geoscientists and students
- Run the simulations, history matches and models a laptop can't
- Open them as apps in a browser tab, with no cloud account and nothing to install
- Keep decks, histories and results on shared disks between runs
- Pay by the hour, and nothing keeps running once you stop
Workflow builders
- Publish from a git repository; PERD builds and pins the image
- Ship the simulator with it: OPM Flow, ResInsight or your own C++, installed at build time
- Add a page and it opens as an app, with charts and a 3-D reservoir grid
- Set your own markup and invoice the people who use it
Build it once, and sell it to the people who need it.
The entry point is a Python function; behind it, anything that installs on the machine: OPM Flow, ResInsight, your own C++. Publish from git and PERD builds the image, pins its interface, opens it as an app and runs it on the machine it asks for. Set your markup, invoice the people who use it, and serve it on your own domain.
Read the docsfrom perd_worker import workflow, WorkflowStreamInput, WorkflowStreamOutput @workflow.bi_di async def history_match( deck: WorkflowStreamInput[bytes], iterations: int = 200 ) -> WorkflowStreamOutput[int, float]: """Match the model to the deck's history; stream the loss.""" case = await simulate(deck) # OPM Flow, installed by setup.sh async for i, loss in fit(case, iterations): yield i, loss
Run the model your laptop can't.
Create an account, top up your wallet and launch your first workspace.