AWS SageMaker: Troubles, Advantages, and Experiments

On Tuesday, the machine learning brand of AQS, AQS SageMaker, announced the release of SageMaker Studio, branded an “IDE for ML.” With its compute-heavy training workloads, machine-learning has been gaining traction. In the growing battle over the public cloud, this might be a decisive factor.

SageMaker’s market share is minimal. By maintenance workloads and streamlining model, SageMaker studio attempts to solve essential pain points for machine-learning and data scientists. Furthermore, its implementation falls short because of long-standing, frequent complaints about AW in general (its sheer complexity and learning curve).

While neglecting features and UX that could relieve from difficulties developers and data scientists face, AWS is embracing a strategy of selling to corporate IT. While the underlying they are realizing, Debugger, Notebooks, and Model Monitors.

Accessing SageMaker studio is a little bit tricky. It is a little bit difficult to set up Studio. Moreover, an existing AWS account cannot log into the new service; instead, you need an SSO (a new AWS single sign-on). Getting a SageMaker Studio session requires understanding the full SSO permissions model – a steep learning curve.

Even though AWS has sought to make machine learning more accessible to customers, many people describe service as technically complex to work with. Hence, this problem is not unique to SageMaker. Many of AWS’s cloud products have the same kind of complexity. It’s competitor, Google Cloud, said that they have a better developer experience. It seems like they are more user-friendly and most caring for the need of professional developers.

SageMaker announced Studio’s new capabilities: Debugger, Model Monitor, Autopilot, Notebooks, and Experiments.

SageMaker: Complexity Over Simplicity

For now, investors do not have to worry about anything. Preferencing complexity over simplicity probably is the right choice.  By that choice, the company is focusing on the large, deep-pocketed corporate IT buyers. These buyers emphasize feature checklists and customizable fine-grained security.

Moreover, it happens at the expense of a steep learning curve and developer friendliness. The complexity opens up the AQS potential of Christensen-Style disruption, but this strategy is rational for now. The advantage of AWS’s sheer size is its more significant economies of scale. Additionally, its ability to support broader offerings, a more extensive certified developer base, and others. Already, this year, we have seen the two b2b companies that circumvented the traditional corporate IT sales path (Slack and Zoom). By forcing the hand of buyers and winning over the minds and hearts of end-users, it has advantages.

SageMaker Notebooks, getting a Python or R environment figuring out and working out how to use a notebook, attempted to solve the most significant barrier for people learning data science.

Single-click Notebooks for the SageMaker environment competes directly against the Microsoft Azure Notebooks and Google Colab in the Notebook-as-a-Service category, which is the companies main goals.

The experiment provides progress reporting capabilities for long-term jobs. It is helpful if a worker has silently crashed in the background of the laptop, or if the user does not know how long a job will continue to run. For cloud-based tasks, GPU-intensive projects, or broad data sets, this experiment’s invention should be a useful addition — it started to work in July 2018. So, data we do not know whether the examination is helpful or not.