Amazon Bedrock Reviews

Amazon Bedrock is used as a bridge between an application and a foundational Gen AI model. It enables me to use publicly available models from companies like Anthropic or Meta through Amazon or AWS. These models are hosted or borrowed by Amazon through APIs, providing a centralized place to utilize different foundational models. 

Bedrock allows comparison of these models for assessing performance and effectiveness for specific use cases. In my projects, Bedrock is used in multiple stages, including data pipeline processes like data cleaning or formatting, sentiment analysis, and creating chatbots for end users. The main strength of Gen AI, which Bedrock leverages, is reasoning, significantly aiding data pipelines.

Bedrock plays a vital role in building a strong foundation for data pipelines by adding reasoning capabilities, often missing from backend workflows. It offers productivity enhancements by providing a playground to compare models and experiment with Gen AI applications. It also aids in clean data preparation and sentiment analysis, leading to better internal workings of applications.

First, Bedrock offers various foundational models in one place. Second, it provides customization options for these models through techniques like fine-tuning, retrieval augmented generation, and continual pre-training, which are innovative features not seen in other managed platforms. Experimentation with Gen AI using Bedrock is notably user-friendly.

I would appreciate a greater focus on agentic Gen AI applications in Bedrock. While Bedrock includes agents in its toolkit, the feature lacks complexity compared to open-source frameworks. 

Additionally, the user interface for the playground could be more refined. While Bedrock is powerful, it lacks markdown formatting features seen in interfaces like ChatGPT.

I have been using the Bedrock solution for approximately six months.

Bedrock manages scalability and reliability effectively as a serverless solution, ensuring AWS handles scalability and security.

Bedrock automatically scales, with AWS handling scalability concerns. Users are responsible for data security at the application level, but AWS provides features to support this.

The AWS technical support was quick to respond, even under a basic support plan, and deserves an eight rating.

Positive

I have worked with LangChain and its agentic framework, LandGraph. However, these do not provide managed services like Bedrock.

Initially, navigating the AWS console for Bedrock was somewhat confusing with multiple clicks needed to reach the required features. However, after regular use, it has become part of my routine and is no longer an issue.

Bedrock offers a very high return on investment due to its cost-efficiency. It uses a pay-for-what-you-use model, allowing experimentation with foundation models at a low cost.

Using Bedrock is inexpensive for experimenting with foundation models compared to managed services from other companies hosting these models.

LangChain and LandGraph, although not providing managed services like Bedrock.

I recommend Bedrock to anyone entering the Gen AI field or considering experimenting with Gen AI. Its cost-effectiveness makes it ideal for experimentation.

I'd rate the solution nine out of ten.

Public Cloud

Amazon Web Services (AWS)

We use Bedrock primarily for its LLM (Large Language Model) capabilities. It serves our needs related to artificial intelligence solutions, including capabilities related to LLM.

Bedrock has been stable and works seamlessly as part of AWS services. Collaborating with it means we can easily integrate solutions however needed, and it's useful as we understand AWS's standard way of operating.

The valuable feature of Bedrock is its flexibility and comprehensiveness in what it's offering, providing parameters that we can change. This differs from Kendra, which doesn't allow parameter adjustments.

One area for improvement is in cost—it tends to be a bit on the higher side, especially for enterprise versions. Furthermore, it lacks certain AI capabilities, such as supporting voice and images.

We have been using Bedrock for two to three months.

Over the past three months, the solution has been absolutely stable with no issues.

Bedrock is scalable. However, there are inherent limitations such as rate per limits and token limits which are standard for LLMs, and we are aware of them.

Technical support from Amazon is satisfying, though Text and Kendra could use improvements. Both are like black boxes with limitations in parameters that could be controlled.

Positive

The initial setup of Bedrock is straightforward, and no significant issues were encountered during deployment.

I can manage the setup independently and do not require an extensive team for installation.

The licensing and overall pricing of Bedrock are competitive compared to other providers like Azure. However, LLM solutions can be expensive when opting for enterprise versions.

The integration is seamless, and given the LLM limitations are known, it allows us to plan and manage accordingly. Awareness of cost implications might be necessary.

I'd rate the solution ten out of ten.

Public Cloud

Amazon Web Services (AWS)

We are using Amazon Bedrock for generative AI-related tasks. We utilize Anthropic Claude LLM to obtain appropriate answers for user questions.

Amazon Bedrock offers an environment where we only pay for the model we use, and AWS handles the scaling. There is no need to maintain any environment, scaling, or DevOps-related tasks. Currently, we use Amazon Bedrock's LLM model to convert questions into answers by providing some context. It provides us with appropriate answers based on that context, which makes data processing tasks efficient.

AWS could add prompt engineering methods to its services. Currently, there are no prompt methods, so we have to experiment on our own. If AWS provided methods, like five or six prompts that yield specific results, it would ease development.

We have been using it for approximately 1.5 to 1.7 years.

I have not experienced any stability issues. Amazon Bedrock is highly scalable with AWS Lambda, AWS Transcribe, and Polly.

I have not faced any scalability issues. It scales well with AWS Lambda, AWS Transcribe, and Polly.

We have not faced the scenario of migrating from a different warehouse. We created a project from scratch on Amazon Bedrock.

After retrieving text from Transcribe, we pass it to Amazon Bedrock using a Lambda. Amazon Bedrock summarizes the text messages and provides appropriate answers, returning questions like the purpose of the meeting or the issues users face.

The pricing depends on the LLM model in use, such as the 1.2 Anthropic Claude. Costs are based on the number of characters obtained in return. It follows a pay-as-you-go model, with different pricing for context and versions like 1.2, 2.3, and 3.1.

Overall, I rate Amazon Bedrock a seven out of ten. It is slightly difficult to integrate with our product. A good knowledge of back-end development is necessary. If users have this, they can proceed. Otherwise, it may not be as user-friendly compared to other services.

Public Cloud

Amazon Web Services (AWS)