Anyscale Platform, Powered by Ray

What do you like best about the product?

What I like best about Anyscale is how it makes it easy to build, deploy, and scale distributed AI and machine learning applications using Ray without the operational complexity of managing infrastructure. The platform provides managed clusters, autoscaling, distributed training, batch inference, and LLM serving in a unified environment, allowing teams to focus on developing AI applications instead of infrastructure management. I also appreciate its excellent performance, seamless cloud integration, and developer-friendly experience. Overall, Anyscale significantly accelerates AI development, improves resource utilization, and enables production-scale machine learning and LLM workloads with minimal operational overhead.

What do you dislike about the product?

One area where Anyscale could improve is providing more granular cost visibility and simpler debugging tools for large distributed workloads. While the platform abstracts much of the infrastructure complexity, optimizing cluster configurations and resource usage for enterprise-scale AI workloads can still require experience. I'd also like to see broader native integrations, richer monitoring dashboards, and more fine-grained controls for workload optimization. Overall, the experience has been very positive, but improved cost transparency, enhanced observability, and additional customization options would make Anyscale even more valuable for teams running production AI applications.

What problems is the product solving and how is that benefiting you?

Anyscale solves the challenge of building and running distributed AI, machine learning, and large-scale data processing applications without the complexity of managing clusters and infrastructure manually. By providing a fully managed Ray platform, it enables teams to scale training, batch inference, online inference, and LLM workloads with automatic resource provisioning and autoscaling. This reduces infrastructure management, improves resource utilization, shortens deployment times, and allows developers to focus on building AI applications instead of operating distributed systems. As a result, it has accelerated AI development, increased engineering productivity, and made production-scale ML and LLM deployments much more efficient.

What do you like best about the product?

You spend more time on the code/workflow and less on the plumbing of provisioning/maintaining distributed resources.

What do you dislike about the product?

ven with managed infrastructure, you still need to think about cluster sizing, concurrency, autoscaling behavior, failure handling, and workload/resource patterns.

What problems is the product solving and how is that benefiting you?

Anyscale is primarily solving the “how do we run Ray reliably at scale?” problem—especially when moving from experimentation to production workloads.

What do you like best about the product?

It helped me create a sandbox for my product, Clara AI, so I could offer it to prospects after the product call.

What do you dislike about the product?

As of now everything is working perfectly smooth and cloud storage for Procol's product is a big plus

What problems is the product solving and how is that benefiting you?

My business requires heavy cloud requirements which is solved by Anyscale

What do you like best about the product?

I like how Anyscale makes it easy to scale distributed AI workloads and simplifies deployment and resource management. It reduces the operational complexity of running machine learning applications. I also found the initial setup to be quite easy, smooth, and pretty good.

What do you dislike about the product?

I find some features and configurations complex for new users, so I think clearer documentation, simpler setup guides, and more intuitive workflows would make Anyscale easier to adapt.

What problems is the product solving and how is that benefiting you?

I use Anyscale for building and deploying machine learning workloads, simplifying scaling, managing distributed computing, and reducing deployment complexity, making it more efficient.

What do you like best about the product?

Anyscale makes it easy to build, deploy, and scale AI/ML workloads with Ray while keeping the developer experience straightforward. I especially like its ability to seamlessly scale distributed workloads, manage GPU resources efficiently, and move from experimentation to production without major infrastructure overhead.

What do you dislike about the product?

The main drawback is the learning curve around Ray and distributed computing concepts, especially for teams new to the ecosystem. Some advanced configurations can also feel complex, and cloud costs can become difficult to predict when running large-scale GPU workloads.

What problems is the product solving and how is that benefiting you?

Anyscale helps simplify the development and deployment of distributed AI/ML workloads by handling infrastructure, scaling, and resource management through Ray. It reduces infrastructure complexity, speeds up experimentation, and makes it easier to move AI workloads from development to production efficiently.

What do you like best about the product?

The app is easy to use, and the service is user-friendly.

What do you dislike about the product?

The app’s cost and service fees of the app.

What problems is the product solving and how is that benefiting you?

It’s helping me develop my music creativity and style, and it also frees up my time so I can be more productive.

What do you like best about the product?

Anyscale removes the DevOps headache of manually managing distributed computing clusters. It handles cluster creation, scheduling, and autoscaling automatically, allowing machine learning engineers to focus on code rather than infrastructure.

What do you dislike about the product?

Anyscale is built entirely around Python and the Ray ecosystem. If your engineering team prefers Kubernetes-native tools, Apache Spark, Databricks, or other orchestration frameworks, Anyscale will feel restrictive.

What problems is the product solving and how is that benefiting you?

The platform helps avoid vendor lock-in by supporting multi-cloud deployments (AWS, GCP, Azure) and a "Bring Your Own Cloud" (BYOC) model. It intelligently manages workload queues and autoscales heterogeneous CPU/GPU clusters to maximize utilization and keep hardware costs down.

What do you like best about the product?

What I liked about Anyscale is how it simplifies running and scaling AI/ML workloads. It provides a flexible environment for developing and deploying Ray-based applications without having to manage as much of the underlying infrastructure. The ability to scale workloads when needed while keeping development and deployment in one place had been particularly useful for our team.

What do you dislike about the product?

The main downside for me was the initial learning curve. If you're new to Ray or distributed computing, it can take some time to understand how everything fits together. The documentation is useful, but more practical examples and simpler guidance for common use cases would make getting started easier.

What problems is the product solving and how is that benefiting you?

The main problem Anyscale helped us with was managing and scaling distributed AI/ML workloads. It reduced some of the infrastucture overhead involved in running Ray applications and makes it easier to scale workloads as requirements grow. This led the team spend more time on the actual ML workloads rather than managing the underlying environment.

What do you like best about the product?

The best thing I like about this platform is its ability to make it easier to develop and scale AI and machine learning workloads just using Ray technology. Also, I like that the same code can move from different interactive workspaces to schedule jobs and protection services, along with minimal changes. Its automatic scaling, workload monitoring and dependency management reduce a lot of infrastructure work. Also, the console provides useful visibility into logs, metrics and other resource usage. Overall, it allowed our team to focus more on the model and applications than cluster management.

What do you dislike about the product?

Their initial learning curve is bit challenging, especially for users who are unfamiliar with ray and distributed computing concepts, understanding and computing configurations on dependencies jobs and services takes some time in it. Their cloud costs can also increase quickly if we are auto-scaling and our ideal resources are not monitored carefully. Some of their advanced configuration and debugging tasks still need strong technical knowledge, and providing beginner-friendly guidance and clearer cost forecasting would definitely improve this platform experience.

What problems is the product solving and how is that benefiting you?

This platform solves the problem of the complexity of running data-intensive AI workloads across its distributed cloud infrastructure. It also removes much of the manual effort involved in configuring clusters and scaling resources, along with deploying the model and other monitoring workloads. This helped us move more quickly from experimentation to protection while using the same ray-based development approach. And another major important thing is their auto scaling and spot instant support can improve infrastructure utilisation and control cost. It ultimately reduces the operational overhead and gives developers more time to improve the actual AI application.

What do you like best about the product?

Anyscale has made scaling training workloads for our customer support assistant much more manageable, distributing compute-intensive tasks across a cluster without needing to manually manage the underlying infrastructure. Being built on Ray meant the underlying distributed computing framework is battle-tested, giving confidence that scaling wouldn't introduce unexpected instability. The interface for managing clusters and jobs is straightforward, letting the team submit and monitor training runs without deep distributed systems expertise. Integration with our existing Python-based training code was smooth, requiring minimal changes to take advantage of distributed execution.

What do you dislike about the product?

The learning curve for effectively using distributed training patterns took real time, especially understanding how to structure code to actually benefit from parallelization rather than just adding overhead. Costs for larger clusters can add up quickly during extended training runs, requiring careful monitoring to avoid leaving resources running unnecessarily. Documentation covers common patterns well, but more advanced or custom distributed workflows occasionally required digging through Ray's broader documentation rather than Anyscale-specific guidance.

What problems is the product solving and how is that benefiting you?

Anyscale has let us scale training for our customer support assistant across a distributed cluster without building and maintaining our own infrastructure for parallel compute. This has significantly sped up training iteration time for larger experiments, letting us test more model configurations in less time than a single-machine setup would allow.