We (Saturn Cloud) published a write-up on the architecture and setup process for running our platform on Nebius. The goal was to make it straightforward for teams to use GPU-backed environments without having to spend much time dealing with Kubernetes operations.
The article walks through how resources are provisioned, how environments are created, how GPU jobs are scheduled, and what abstractions we use to keep the system flexible while hiding most of the complexity of the underlying cluster. It also includes some of the design decisions we made along the way and a few of the tradeoffs we ran into.
Since we built the system, I’m happy to answer questions about the architecture, decisions, limitations, or areas we are still iterating on.
This has not been true for years. My google ads account was blocked, I had several google engineer friends create internal tickets for my case, and none of it worked out.
Can anyone comment on external displays? The last time I tried this was about 5 years ago. I had a Samsung chromebook which was a great little device, but the font rendering connected to an external display was too terrible.
At saturncloud.io we focus on helping data science teams move to the cloud (either in your AWS account or ours). One component we provide is hosted jupyter lab (along with SSH access so you can use PyCharm/VSCode or anything else.
Hobbyists often use our free tier (30 hrs of GPU per month) or pay for more compute ($0.69/hour) to do a variety of GPU workloads.
We do prohibit things like DDOS attacks and cryptocurrency mining.
The area that we've seen Dask shine the most is ML. This is because the Python ecosystem has more advantages when it comes to ML and there is better GPU support.
Disclaimer: We produced those benchmark. I'm a founder of Saturn (https://www.saturncloud.io/) and we focus on providing Databricks-like capabilities with Jupyter + Dask + Prefect, so I definitely have strong feelings in this area.
For some context on why this matters, if you're writing a library (like sklearn) and you want to support multiple array types, you might need to do stuff like
if isinstance(x, ndarray):
...
elif isinstance(x, other_array):
...
In the most ideal case having the standard means that scientific libraries can support all conforming implementations by default. Then sklearn would automatically support cupy/numpy/dask/jax/mxnet/pytorch/tensorflow arrays. Multiply that by all the scientific libraries and the effect is pretty profound.
Just to set some context - this is being driven by Quansight which is led by Travis Oliphant, who united Numeric and Numarray to create NumPy many years ago which lead to the entire python data science ecosystem as we know it. Yes it's ambitious, and they may not succeed, but it's not their first barbecue. Unifying communities and APIs is something they're good at.
We’re a team of 10 and we’ve also found that Cadence replaces our need for stand up meetings. In general, I’ve found that trusting my team to get their work done and unblock themselves has resulted in fewer synchronous meetings
Saturn Cloud | Remote (North/South america) | Senior Software Engineer | Full Time
Saturn Cloud is DataBricks for Dask. We're building an integrated data science platform leveraging Jupyter, Prefect and Dask. Our code base is in Python, Vue/TypeScript, and lots of k8s. We also have some roles that are 50% allocated for contributing to the scientific python ecosystem (mostly Dask, but also sklearn/pandas + friends). We only have 6 engineers right now, so you'll make a big impact.
Right now we're hiring people who can build Python web backends and are somewhat comfortable building UIs. Understanding/empathy of data scientists is a plus, as is understanding of devops/aws/azure/k8s.
The team is fully remote, but we keep to similar time zones. For most regions, our salaries are quite competitive.
Our interview process: Phone Screen, a few 30 minute chats with team members (to give you a sense of what we're like, and help you figure out if you want to work here), 2 hour take home project, 2 hour pair programming session. The take home project is time boxed so that we don't burn your time.
Sorry to approach here, but don't know how else to reach you. I'm looking for remote Vue devs, maybe we can work together? If you'd like to talk, please reach out to [email protected]
There isn't a good way to know for sure that something is going to work, but there is alot of research you can do that will give you good information.
1. Cold email/linkedin message then speak to 100 potential customers. Attempt presales. You don't actually have to get presales, but that process should give you a good indication for willingness to pay / desired functionality, how people are currently working on the problem, and what language they use to describe their pain (quoting that back to customers is good marketing copy)
2. Build a landing page, run facebook / google ads, find out how much it costs to acquire an email address. Your real CAC will be higher of course, but this is a good starting point.
I'm a founder of https://www.saturncloud.io/. I started with #2, and found it cost 70 cents in advertising revenue to get a click, ~7$ for an email address, and ~$100-$200 to get someone to signup. When we started we were attempting a low touch sales model.
Since then we've started focusing on enterprise sales, because the willingness to pay is much higher. We did #1 when moving towards enterprise sales, but should have done that in the beginning.