Generative AI Lab: No-code Model Training and Tuning

Generative AI Lab is offered as a pay-as-you-go solution on cloud marketplaces. The software subscription includes:

• GPU support for enhanced speed of pre-annotation and training,
• Visual Document Understanding features as well as support for training Visual Document Classification and Visual NER models,
• Preannotation with LLM prompts and Zero-Shot prompts,
• 2,000+ Healthcare tuned embeddings and AI models for classification, entity extraction, entity resolution, relation extraction, and assertion status detection,
• Preannotation via rules, models, or prompts,
• Premium support.

No. There is no limitation imposed on the number of projects, users, or documents that can be annotated within the Generative AI Lab. No limitation on the number of models, prompts or rules you can define, test, train or tune with the Generative AI Lab.

Also, there is no limitation on the number of pre-annotations you run or on the number of models you can train.

No, installation through the marketplaces will generate a cost, and using an on-premise deployment will require a license key. Contact us at support@johnsnowlabs.com for on-premise deployments.

Installation instructions are available here: https://nlp.johnsnowlabs.com/docs/en/alab/install

Yes. The Generative AI Lab will replace the NLP Lab products on the AWS and Azure Marketplaces. You can continue to use your existing subscription until the end of 2024, when we will end support for these products in these marketplaces.

We currently offer the product on AWS Marketplace and Azure Marketplace.

Yes. The Generative AI Lab can be used in high-compliance industries like healthcare, life science, finance, and insurance where on-premise deployments are common.

Most single-machine, and Kubernetes distributions are supported.

Yes. Make sure to allocate enough memory & compute power for your use case.

This depends heavily on your use case. The minimal required configuration for on-premise deployments is an 8-Core CPU, 32GB RAM of memory, and 512 GB of SSD storage.

The recommended configuration to support model training and AI-assisted pre-annotations for a team building or validating text models is a 16-Core CPU, 64 GB of memory, and 512 GB of SSD storage. The recommended configuration for teams using Document Understanding features and Visual model training is a 4-GPU instance with 48 CPU cores and 192 GB RAM, equivalent to g4dn.12xlarge AWS instances.

The software price is calculated based on usage and on the type of server where it is deployed.

Usage of Visual Document Understanding and of Healthcare features are charged based on consumption per vCPU per hour.

Charges are reflected on your AWS or Azure bill, and billed through your cloud provider.

Yes! Please email us to describe your situation and needs.

No. The software is designed to be installed and operated entirely within your own infrastructure. It is built with privacy and data sovereignty in mind, ensuring that it does not transmit any data or results outside of your controlled environment.

This architecture guarantees that your data remains within your jurisdiction, providing you with full control over its security and privacy.

You do. We will never even see them.

Our software is engineered specifically for environments that require strict compliance and robust security measures. It operates directly within your infrastructure, ensuring that all data processing occurs locally.

This means your data, including any Protected Health Information (PHI) or Personally Identifiable Information (PII), remains within your control at all times and is never sent to John Snow Labs.

We also provide the option to integrate with third-party Large Language Model (LLM) services, such as OpenAI, to leverage features like prompt-based pre-annotation and synthetic data generation. Should you decide to utilize these functionalities, the responsibility for implementing appropriate safeguards to securely and privately share PHI data with these external services rests with you.

It is crucial to ensure that any data sharing complies with your organization's privacy policies and relevant regulatory requirements to maintain the confidentiality and integrity of sensitive information.

Generative AI Lab does necessitate an active Internet connection for its operation. This requirement serves two primary purposes:

• Metering Usage Collection: An Internet connection is essential for transmitting usage data. This allows for accurate metering of your usage of the platform, ensuring that you are billed correctly based on your actual usage.
• Model and Pipeline Downloads: In order to utilize the advanced features of Generative AI Lab, such as pre-annotating documents, an Internet connection is required to download the necessary models and pipelines from the NLP Models Hub. This ensures that you have access to the latest and most efficient tools for your data processing needs.

We understand the importance of consistent and reliable access to these resources and functionalities, and an Internet connection ensures that Generative AI Lab can deliver its full capabilities to enhance your projects.

Same business day 8x5 support is included with all subscriptions. We can also provide 24x7 support for production systems - please email us if you require it.

Here: https://nlp.johnsnowlabs.com/docs/en/alab/quickstart

Each license includes the software libraries in all supported languages, the pre-trained models that are included with it, premium support, and all updates to the software & models that are released during the subscription period.

Two licensing options are available for the John Snow Labs libraries:

• Pay-as-you-go subscriptions, based on the type & amount of compute time being used.
• An annual subscription, payable once a year in advance.

No. There is no limitation imposed on the number of characters, words or documents processed with the free or paid libraries.

The pay-as-you-go license allows parallel use of the software on multiple machine and by multiple users in parallel. All usage is metered and summed in your monthly invoice.

Pay-as-you-go free trials automatically convert to paid pay-as-you-go licenses at the end of the free trial. Continue to use the same license key, and you will be charged based on your consumption.

For annual subscriptions, the software will stop processing documents - for both training and inference. If you choose to buy a license, we will provide you with a new license key. Otherwise, you must uninstall the software. In any case, data you have already processed is yours to keep.

Python.

We officially support AWS, Azure, GCP, OCI, Databricks, Snowflake, Cloudera, Colab, Kaggle, Docker and Kubernetes.

Yes. John Snow Labs NLP is used heavily in high-compliance industries like healthcare, life science, finance, and insurance where on-premise deployments are common. Most single-machine and Kubernetes distributions are supported.

Yes. Make sure to allocate enough memory & compute power for your use case.

Yes. Make sure to allocate enough memory & compute power for your use case.

This depends heavily on your use case. For training custom models based on the BERT family of embeddings, at least 8 cores and 64GB of memory are recommended. For inference, as little as 1 core and 16GB may be enough. Using GPU’s will usually provide faster execution at a higher cost.

Pay-as-you-go licenses are charged based on consumption, per vCPU per hour.

For annual subscriptions, the cost depends on which edition you need (Healthcare, Visual, Finance or/and Legal), the level of support (8x5 or 24x7), and the number of licenses you need. Please email us with those details at sales@johnsnowlabs.com to get a custom quote.

Online bank transfers (ACH or wire), checks, and all major credit cards. You may be able to use AWS or/and Azure credits, depending on the terms of your cloud credits.

Yes! Please email us to describe your situation and needs.

No. You install and run the software on your infrastructure. The software does not “call home” and no data or results are sent to John Snow Labs.

You do. We will never even see them.

This is not a SaaS solution - instead, you run the software on your infrastructure. Nothing ever gets sent to John Snow Labs or another third party. John Snow Labs NLP is designed for high-compliance, locked-down environments.

No, after an initial installation & downloading of pre-trained models.

Yes, an internet connection is necessary to establish the validity of your license key when starting your Spark session. For airgap deployments contact our sales team at sales@johnsnowlabs.com to accommodate your specific needs.

No.

Yes. John Snow Labs NLP is designed to enable you to train & tune your own models for most tasks.

Yes. Here’s an example: https://github.com/JohnSnowLabs/spark-nlp-workshop/blob/master/tutorials/Certification_Trainings/Public/4.NERDL_Training.ipynb, section NERDL Model.

Yes. Here’s an example: https://www.youtube.com/watch?v=mfd_7a7HHwE&t=4s

Yes. Here’s an example: https://towardsdatascience.com/text-classification-in-spark-nlp-with-bert-and-universal-sentence-encoders-e644d618ca32

Absolutely! Here’s an example: https://github.com/JohnSnowLabs/spark-nlp-workshop/blob/master/tutorials/Certification_Trainings/Public/2.Text_Preprocessing_with_SparkNLP_Annotators_Transformers.ipynb

The full list is available here. Expect the list to keep growing over time.

Email support@johnsnowlabs.com, call us at +1-302-786-5227, or start a chat on spark-nlp.slack.com. Paying customers get a private Slack channel, so that you can ask your questions privately.

Same business day 8x5 support is included with all subscriptions. We can also provide 24x7 support for production systems – please email us if you require it.

Yes. The Demos and Notebooks Hub includes links to runnable Google Colab notebooks in Python.

Here: nlp.johnsnowlabs.com/docs

A private LLM endpoint is a secure, dedicated API endpoint that provides access to pre-trained Large Language Models (LLM) or Natural Language Processing (NLP) models. These endpoints are designed for use within a specific organization's security perimeter (that is, within your VPN or firewall), ensuring data privacy, security, and compliance with regulatory requirements.

No. There is no limitation imposed on the number of characters, words or documents processed with the private AI endpoints.

Free trials subscriptions for private LLM endpoints automatically convert to paid pay-as-you-go licenses at the end of the free trials. You can continue to use the private endpoints and you will be charged based on your consumption.

John Snow Labs' private endpoints can be used with a variety of programming languages that support RESTful API calls. For instance, Python, Java, Scala, R, .NET (C#), JavaScript (Node.js), Ruby, Go, etc. Each of these languages can leverage RESTful API calls to interact with the private endpoints, ensuring that you can integrate the LLM & NLP capabilities into your applications regardless of the technology stack you are using.

John Snow Labs' Private Endpoints are available on the following platforms:

• AWS Sagemaker - AWS Sagemaker provides a robust environment for deploying machine learning models and integrating with John Snow Labs' NLP capabilities.
• Databricks Serve - Databricks Serve offers a seamless platform for deploying machine learning models and serving LLM & NLP models from John Snow Labs, leveraging the powerful capabilities of the Databricks environment.
• Snowflake Native Apps Marketplace - Snowflake Native Apps Marketplace allows for easy integration and deployment of John Snow Labs' LLM & NLP models within the Snowflake ecosystem, providing data processing and analytical capabilities directly within Snowflake.

Private endpoints are charged based on vCPU consumption per hour, based on the infrastructure used for deployment. Check pricing details on the marketplace pages of the private endpoints.

The payment methods for each marketplace are as follows:

• On AWS Sagemaker, the consumption generated by private endpoints is included in your AWS bill.
• Databricks Serve endpoints require the purchase of a PAYG license key from my.johnsnowlabs.com. At subscription time you need to provide your credit card details and payments are facilitated via Stripe.
• On Snowflake Native Apps marketplace, the consumption generated by the private endpoints is also included in your Snowflake bill.

No. We currently do not offer academic discounts for private endpoints.

No. You install and run the software on your infrastructure. The software does not “call home” and no data or results are sent to John Snow Labs.

This is not a SaaS solution - instead, you run the private endpoints as part of your VPN. Nothing ever gets sent to John Snow Labs or another third party. John Snow Labs NLP is designed for high-compliance, locked-down environments.

No.

Model training features are not offered as private endpoints. To train or tune your own medical language models, use the Medical Language Model library subscriptions. This software library is designed to support customize pipelines, training including transfer learning, integraiton with MLOps tools like MLFlow, logging and evaluation of trained models, and more.

Email support@johnsnowlabs.com, call us at +1-302-786-5227, or start a chat on spark-nlp.slack.com. Paying customers get a private Slack channel, so that you can ask your questions privately.

Same business day 8x5 support is included with all subscriptions. We can also provide 24x7 support for production systems - please email us if you require it.

Here: https://sparknlp.org/docs/en/quickstart

The Medical Chatbot Professional version is provided as a Software-as-a-Service (SAAS) offering for healthcare professionals and researchers. This package includes a suite of features designed for optimal user experience and accuracy in medical assistance:

• Chat Interface: An intuitive and user-friendly chat interface facilitates seamless interactions, ensuring ease of use for all users.
• Healthcare-Specific Large Language Model: Our proprietary Large Language Model is fine-tuned specifically for healthcare applications, offering precise and contextually relevant responses.
• Daily Medical Knowledge Updates: The platform is regularly updated with the latest medical knowledge, ensuring that the information provided is current and accurate.
• Enhanced Medical Conversations: Designed to simulate natural and informative conversations, this feature allows for efficient patient engagement and information dissemination.
• Smart Ranking of References: This functionality prioritizes the most relevant and authoritative references, enabling users to access credible information swiftly.
• Robust Safeguards: Built-in safeguards are designed to maintain the integrity of the information provided, ensuring reliability and trustworthiness.
• Hallucination Prevention: Advanced mechanisms are in place to minimize the risk of generating inaccurate or misleading information, enhancing the platform's reliability.
• Searchable Bookmarks and Responses: This feature enables efficient retrieval of previously discussed topics and answers, streamlining the information search process.

The Enterprise edition of the Medical Chatbot is tailored for private deployment, focusing on the specific needs of your team and data. It offers a suite of features designed for organizations seeking the highest level of customization, security, and scalability:

• Private On-Premise Deployment: Ensures that the solution is installed and operated within your own infrastructure, providing complete control over data privacy and security.
• Unlimited Custom Knowledge Bases: Allows for the integration of custom knowledge bases, ensuring that your specific informational needs and nuances are addressed.
• Direct Integration with Your Data Sources: Facilitates seamless connection to your existing data sources, enhancing the relevancy and personalization of the information provided.
• Customizable Brand Voice and Safeguards: Offers the ability to tailor the solution’s interaction style to match your brand identity, alongside advanced safeguards to maintain data integrity and confidentiality.
• Unlimited Users and Groups: Supports an unrestricted number of users and group configurations, enabling wide-scale deployment across your organization without additional constraints.
• Role-Based Access Management: Implements granular access controls, ensuring that users only have access to appropriate levels of information and functionality based on their roles.
• Single Sign-On (SSO) Capability: Simplifies user authentication by integrating with your existing single sign-on systems, streamlining access for users.
• API Access: Provides comprehensive API access, allowing for custom integrations and enhancements to fit your unique operational workflows.
• Scalable Architecture: Designed with scalability in mind, this edition is capable of accommodating growth and evolving requirements, ensuring that the solution remains effective as your organization expands.

No. Professional accounts are for individual users. If you want your team to use the Medical Chatbot you must create individual accounts for all team members. Alternatively, you can opt for the Enterprise edition to deploy on your premises.

The monthly subscription costs $49.

All major credit cards.

The cost of the Enterprise Medical Chatbot is calculated based on your specific needs and the level of customization necessary. Please email us to describe your situation and needs.

You do not need any special setup for this. You can start chatting from your browser.

Yes. The Medical Chatbot can be used in high-compliance industries like healthcare, life science, finance, and insurance where on-premise deployments are common.

We recommend utilizing NVIDIA A100 80 GB instances with 8 GPUs.

When utilizing the Professional (SaaS) Edition, your queries are processed by our software. It is imperative to exercise caution to ensure that Protected Health Information (PHI) is never sent or included in chats with the Professional Edition.

In contrast, the Enterprise Edition does not share any data with John Snow Labs or any other third party, and is designed so that you can use it to analyze and discuss PHI and other sensitive content.

For analyzing PHI or PII data, use the Enterprise Medical Chatbot, designed for on-premise deployment with advanced security features. It employs rigorous authentication, authorization, and granular access control systems, allowing administrators to precisely manage who can access sensitive data.

It integrates seamlessly with LDAP and Active Directory for streamlined user management and supports standard security protocols like OpenID Connect and OAuth 2.0, ensuring compliance with best practices. Additionally, it offers social login capabilities for ease of use without compromising security.

We provide 24x7 support for Enterprise deployments.

Here: https://nlp.johnsnowlabs.com/docs/en/chatbot/healthcare_gpt