Modal

What do you like best about the product?

Modal provides a seamless infrastructure-as-code experience that makes running cloud compute feel like local Python development. Decorating functions let's you execute code on remote GPUs or CPUs instantly without writing Dockerfiles or managing Kubernetes clusters. Sub-second cold starts, GPU memory snapshotting, per-second billing, and automatic scaling to zero eliminate idle server costs and simplify running AI inference and background batch jobs.

What do you dislike about the product?

The primary limitation is it's heavy optimization around Python, making it less suited for microservices built in Node.js, Go, or Rust. Additionally, reliance on modal's custom SDK introduces platform vendor lock-in, and 24/7 continuous workloads can become expensive compared to reserved cloud instances.

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

It eliminates the heavy operational overhead of maintaining AI and batch compute infrastructure. Backend developers can deploy scalable ML pipelines, background tasks, and code execution sandboxes directly from standard python code in hours rather than spending days configuring servers and cluster scaling.

What do you like best about the product?

I like how easy it is to run code on cloud computers. You don’t have to deal with cloud settings or go through long setup guides. You just add a tag to your Python code and it runs in the cloud automatically. It also scales down when you’re done so you only pay for the seconds your code is running.

What do you dislike about the product?

It's built for tasks and short jobs so it’s not a good fit if you need a server that stays on all the time. Also if you keep running tasks nonstop for a whole month the, per-second billing can add up and become expensive.

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

It removes all the trouble of setting up hardware or complicated server systems. You can write your script on your computer and send it straight to powerful cloud computers with just one command.

What do you like best about the product?

As an AI engineer on our claims platform, we handle commercial retail loss cases where adjusters upload massive warehouse inventory spreadsheets-often exceeding 100,000 individual line items. We cannot process these sequentially in our FastAPI backend without triggering massive timeouts. I appreciate how Modal Labs allows us to instantly fan out these heavy inference workloads. Using Modal's Python SDK, we chunk the massive CSV files and simply wrap our extraction logic in their decorators to instantly spin up hundreds of serverless GPU containers in parallel. Instead of bottlenecking our API by generating embeddings and parsing SKUs sequentially, we use Modal to run our heavy PyTorch extraction models across tens of thousands of line items simultaneously. Because of Modal’s custom container runtime and sub-second cold starts, it aggregates the parallel results and returns clean JSON to our LangChain orchestration in minutes instead of hours.

What do you dislike about the product?

Because the execution and parallelization happen entirely on Modal's cloud infrastructure, true offline local debugging of our GPU worker functions isn't possible. We are completely reliant on their CLI and a stable internet connection to test code changes. Additionally, while their container snapshotting is incredibly fast for warm starts, if we introduce a massive new PyTorch dependency or a proprietary embedding weight into our image layer, the initial remote build process introduces a temporary pipeline delay in the CI/CD pipeline before the cache is warm.

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

Modal completely solves the compute and memory bottlenecks of processing massive commercial inventory files. Before Modal, parsing, mapping, and pricing a 100,000 line-item warehouse claim using standard Azure virtual machines took hours and constantly risked memory crashes on our FastAPI endpoints. By using Modal for horizontal scaling and parallel inference, we distribute the workload across hundreds of temporary GPUs. This drastically accelerates the settlement time for large commercial losses and optimizes our compute budget, as we are only billed for the exact seconds the parallel GPUs are active.

What do you like best about the product?

Honestly, the biggest thing is how little setup it needs. Our IT team writes Python, adds it as a decorator, and the function runs in the cloud. No Docker files to babysit, no Kubernetes, no server config. Containers start quickly, and when we need a GPU we just ask for one in the code instead of raising a request and waiting. Logs stream back live, which makes debugging much less painful. Billing is per second, so we only pay when something is actually running. For a small team like ours without a dedicated platform engineer, that matters a lot.

What do you dislike about the product?

Because it uses its own decorators and SDK, our code is tied to modal. Moving to another platform later would mean some refactoring.

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

We use it for batch jobs, background processing, and running AI/ML workloads without managing our own servers. It has saved our IT team a lot of time on deployment and scaling. Things that used to take days to set up can now be tested and deployed in a few hours. The team spends more time building and less time on infrastructure.

What do you like best about the product?

modal has honestly been a good game changer for running and scaling python workloads and ml models . i am as developer experience is super smooth . modal is highly recommend it if you want to focus on code rather than ops..

What do you dislike about the product?

sometimes it is hard to keep track of the billing when running heavy GPU tasks and the guides could use a few more examples for complex setups....

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

it helps us run python and ai models on heavy GPUs without dealing with complicated server setups. It saves us a lot of time on infrastructure and lets us launch code instantly...

What do you like best about the product?

I really like how simple Modal Labs makes running AI workloads. The platform handles a lot of the infrastructure and scaling, allowing me to focus more on building and testing models. I also appreciate the flexibility of accessing GPU compute when needed without having to manage dedicated hardware.

What do you dislike about the product?

One area that could be improved is the learning curve for new users. Some of the setup and configuration can take time to understand, especially if you are not very familiar with cloud infrastructure. More beginner-friendly documentation and examples would make it easier to get started. For beginners, the initial setup and understanding how to structure deployment can take some time. It would be helpful to have more step-by-step guides, clearer examples for common use cases, and simpler explanations of configuration options. A more guided onboarding experience would make the platform easier to navigate for first-time users.

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

Modal Labs simplifies AI workload management by handling infrastructure and scaling, making it easier to deploy and test models without managing servers. The flexibility in accessing GPU compute helps me focus more on development than on maintenance.

What do you like best about the product?

It's amazing and helps alot really ans literally

What do you dislike about the product?

Sometimes it is very very confusing and doesn't help

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

Optimising my time

What do you like best about the product?

I like that Modal makes it easy to run and deploy Python workloads without having to manage the underlying infrastructure. It is especially useful for running AI and compute-heavy tasks when I need extra resources.

What do you dislike about the product?

It can take some time to understand the platform and its deployment model. Debugging can also be a little difficult when something goes wrong in a remote environment.

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

It simplifies running compute-heavy and AI workloads without needing to manage servers manually. This helps us run jobs on demand, scale resources when needed, and spend less time managing infrastructure.

What do you like best about the product?

Modal provider is best-in-class for developer experience when working with Python. I used it to host my own STT and TTS models, and the overall workflow felt smooth and straightforward. On top of that, their free tier is very generous—like, very, very generous. look guys if you are looking for provider to host your agent or model like wisper and other then its really good option to use model i will higly suggest as its native support for paython deployment can give trouble on paid plan (might be withme bit still)

What do you dislike about the product?

Deployment can be a bit unpredictable sometimes. For example, it failed on me once when I tried to deploy it.

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

As I said previously, it failed once during deployment.

What do you like best about the product?

What I like best about this platform is how easily they let us scale AI workloads at random without even managing any servers or Kubernetes. Its own Python-first approach also makes deploying inference, training, and other batch processing jobs feel very much similar to writing local code. And also GPU selection, container dependencies and other auto-scaling can also be defined within the application. Faster container startup and access to different GPU types are particularly useful when workloads fluctuate. They pay for usage model also helps us to avoid spending on ideal infrastructure.

What do you dislike about the product?

This platform has a bit of a learning curve because its code-first deployment model is kind of different from the conventional cloud infrastructure setup. Also, debugging distributed jobs and optimising its cold start behaviour might require familiarity with model-specific concepts. And their cost may also be bit difficult to predict when highly parallel GPU workloads scale suddenly; this will create bit of confusion in the middle period, as some of their advanced capabilities and other enterprise controls are limited to higher price plans. And more visual configurations and other core cost forecasting tools would make the platform easier for non-technical teams.

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

This platform helps us to remove the infrastructure which is involved in provisioning GPUs and building other containers, along with scaling AI applications. We can also use it for model inference, fine-tuning, and data processing for secure code execution without even maintaining any dedicated computing clusters. And also applications which are automatically scaled according to demand and can scale down when unused. By reducing the initial GPU expenses, this platform is saving lots of cost involved at the initial setup. Overall, this platform allowed our team to test ideas and move models into production much faster. It also reduces the operational overhead while giving our developers more time to focus on the actual AI product.