Mystic AI - Product Update, Feb '24

Hello! Hope everyone's having a great week. We have had an extremely busy few weeks at Mystic. We've been hard at work: launching our v4 API and introducing an entirely new service. We have been waiting to share these new improvements, so here goes:

We listened to user issues with serverless APIs and built: Bring Your Own Cloud (BYOC)

At Mystic AI, we're committed to solving the challenges our users face with ML deployments. And to do so, we listen to what our customers have to say.?

A recurring theme among companies was the initial satisfaction with serverless solutions, which quickly turned to frustration due to inconsistencies in performance, unexpected downtime, and a lack of full control over infrastructure and cost management. Unexpected skyrocketing costs, in particular, was a very common concern. Having identified all these issues, we have developed Mystic BYOC.?

Features include:

• One-click-deploy from our model hub: When a new model gets released on our model-hub, you can migrate it straight away using just a few clicks.
• Use your existing cloud credits: Because the cluster is deployed on your own cloud account, you can pay for runtime using existing cloud credits.
• Granular customisability: You can select your cluster to be running on spot instances or on-demand.
• We provide an optimised to your usage auto-scaler, but if you prefer more control, you can tune it to your preferences;a. Control min. and max. number of GPUs.b. Warmup endpoint. i.e, request a pipeline to get cached on a GPU.c. Cooldown endpoint, i.e, remove a pipeline from cache.d. Windows of traffic, i.e, pay more attention to specific traffic patterns. And much more...
• Transparent pricing structure: And most importantly, BYOC serves as your comprehensive MLOps team on a platform for a monthly flat-fee. So you get absolute control over your infrastructure and your costs with fixed overheads

Sounds interesting? You can try it out yourself now, find the link in the comments!

Happy coding!