We're pretty excited about today's release. This emoji sums it up well: 🎉. Let me try to get you pumped up, too.
tl;dr We're launching a Teams plan on FloydHub today. You can now have a secure, centralized hub for your entire data science team's model development, training, and deployment pipeline, along with roles-based permissions, consolidated billing, and even more collaboration magic.
For the last year or so, ever since our surprise launch on ProductHunt, we've been shipping features that help data scientists everywhere get up and running faster on GPU-powered deep learning machines. There's a whole lot of fiddly-stuff that goes into configuring a deep learning machine, from drivers to frameworks to data, that just plain sucks.
Maybe you've tried this yourself - you probably found the perfect explainer blog-post that walks you through setting up AWS AMIs and VPCs and whatnot. Then, the next thing you know, it's three days later and you still haven't gotten a Jupyter Notebook running and they just released a new version of TensorFlow and you've accidentally spent next month's rent on a cloud instance that you had no idea you even turned on.
FloydHub was built to remove these headaches. With one simple
floyd run command, you're up and running on a fully-configured deep learning machine in just a few seconds. This tweet sums up the experience nicely:
@FloydHub_ feels like magic.— Fast ForWard ►► (@FFW)
But it's this tweet that really speaks to our long-term vision for FloydHub:
@FloydHub_ this is so great, we've been spending a ton of time streamlining our workflow we could be spending improving our models 👍🏽— David (@boxcardavid)
Our prime directive here at FloydHub is to help data scientists be more productive. And with today's release of the FloydHub for Teams plan, we're introducing a new collaborative workflow for data science teams of all shapes and sizes.
Playing nicely with others
We've spoken to data science teams at companies all over the globe -- asking them about their challenges and problems (and hopes and dreams). Here's a laundry list of what we've heard:
- Reproducibility: Everything from data munging to hyperparameter sweeping to model deployment is hand-rolled and difficult to share amongst a team.
- Visibility: Lack of insight into what teammates are working on, right now and historically.
- Productivity: The wheel is constantly being recreated, during every step of the data science process. The bus factor is a real thing.
- Data wrangling: Data is still hard, and data engineering teams will have an increasingly important role going forward.
The FloydHub for Teams plan is here to help - offering a collaborative, centralized hub for all data science experimentation, training, and deployment.
The gooey details
By creating a team on FloydHub, you and your data science team will have access to:
- Centralized, secure hub for all projects and datasets
- Team management, with roles-based permissions
- Usage tracking, across team jobs
- Consolidated billing
- High priority customer support
Sure, but what's it really like?
As you might expect, we've been beta testing the Teams plan over the last few weeks with a few customers.
Collaborating was more of a challenge for us than training complicated models. We often found it difficult to make time and effort for code reviews and dataset sharing. It was like working in an invisible silo.
With FloydHub, things have changed. No more dataset requests, messed up version control or knowledge-sharing bottleneck. My teammates' latest codes and datasets are in just a few clicks away. Plus, no need to buy expensive GPUs for each and everyone.
That's it. We know that data science is challenging, but it doesn't have to be frustrating.
FloydHub powers more productive data science teams by letting them focus on what they do best - the science. Meanwhile, we'll be here to manage everything else. With FloydHub, there's no more reinventing the wheel - transparency and reproducibility come out of the box. With the full experimentation audit trail of FloydHub, your new team members will be onboarded faster and become productive sooner, leading to better results for your entire product and team.
You can sign up for the FloydHub Teams plan today, right now. Just create a team, invite your colleagues, and get to work!