Compute should accelerate worknot become the work
Built for the teams that depend on compute — scientists, engineers and developers running workflows across cloud, laptops, GPU servers, HPC and edge.Our goal: make every compute resource easier to use, operate and trust.
Why we arebuilding dxflow
One project can span a laptop, an EC2 instance, a GPU workstation and an HPC cluster — each with its own tools, access methods and operational burden.
dxflow adds a common layer across them: one way to run workflows, manage files, access shells, monitor systems and connect services. It runs directly on the machines you already use instead of pulling data and control into a central system.
Technical work should not be constrained by where it runs
Four decisions that shape everything we build.
Infrastructure-agnostic
Cloud, on-premise, HPC, local, and edge — the same experience everywhere.
Workflow-first
Workloads, containers, artifacts, logs, and shells from one platform.
Built for existing systems
Works with the runtimes you already use: Docker, Podman, Singularity, and Apptainer.
Open and portable
Keep control of your workloads, data, and infrastructure. No lock-in.
From orchestration to intelligent operations
The next platforms will not just run workflows — they will help teams operate them, with people still in the loop.
Unified compute operations
Connect machines, run containerized workflows, manage files, access shells, and expose services securely.
AI-assisted operations
An agent that builds workflows from plain language, runs them, reads the logs, and fixes what failed — on Bedrock, Claude, Gemini or OpenAI.
Permissioned AI operators
Sessions scope to a workflow and answer to per-key permissions today; standing instructions and fleet-wide operators come next.
What we believe
Four principles that decide what we build — and what we leave out.
Compute should be accessible wherever it lives.
A laptop, cloud instance, GPU server, and HPC cluster should not feel like different products.
Reproducibility should be built in.
Teams should be able to understand, rerun, monitor, and share the work they perform.
AI should be useful, transparent, and permissioned.
AI should help operate complex systems within clear boundaries and human control.
Infrastructure should support innovation, not slow it down.
Less time configuring systems, more time solving meaningful problems.