Generative AI Lab

Generative AI Lab, developed by John Snow Labs, is a no-code platform for building, validating, and fine-tuning NLP and AI models through Human-in-the-Loop (HITL) workflows. It is designed for enterprises in healthcare, life sciences, finance, and legal sectors that need to meet regulatory standards such as HIPAA and GDPR without requiring ML engineering expertise. The platform supports annotation of text, PDFs, images, HTML, audio, and video, and can be deployed on AWS, Azure, or on-premise — including air-gapped environments.

The platform has been used by 500+ enterprise customers, including healthcare systems, pharma companies, payers, and IT organizations, accumulating over 7.5 million annotation hours. It replaces the earlier NLP Lab product on AWS and Azure Marketplaces.

Document Annotation Capabilities

• High-throughput annotation of thousands of documents daily through a high-productivity UI with keyboard shortcuts and pre-annotations.
• Shareable annotation guidelines and reviewer comments to maintain consistency across teams.
• Consensus analysis tools to help annotators reach agreement efficiently.
• Support for annotating text, PDFs, images (PNG, JPEG), HTML, audio, and video content types.
• Document splitting into sentences, paragraphs, or pages for more precise labeling of long documents.
• Adaptive taxonomy customization per document section.

AI-Powered and LLM-Assisted Labeling

• Automatic text pre-labeling to reduce manual annotation effort.
• Zero-shot prompt support for classification, named entity recognition (NER), and relation extraction.
• OpenAI integration for text classification and NER with secure data transfer.
• LLM-based bootstrapping of training examples for classification, NER, and relation detection.
• Medical terminology coding that automatically resolves ICD-10, LOINC, CPT, SNOMED CT, RxNorm, and MeSH codes during annotation.
• API integration to push and pull data from EHRs, data lakes, or ML pipelines.

Model Training, Tuning, and Testing

• Train-as-you-go with active learning, allowing models to improve incrementally as annotations are added.
• Transfer learning from available pre-trained models.
• Built-in performance metrics for comparing model versions.
• Automated model improvement based on annotated data.
• Access to 2,600+ healthcare-tuned embeddings and AI models covering classification, entity extraction, entity resolution, relation extraction, and assertion status detection.
• Live playground for experimenting with models, rules, and prompts on custom data.
• Private hub for storing and managing models, rules, prompts, and medical terminologies within a team.

De-Identification of Patient Data

• Automated detection and removal of Protected Health Information (PHI) from medical records.
• Supports both obfuscation and masking approaches.
• Entity-level configuration for granular control over what is de-identified.
• Consistent de-identification applied throughout an entire task.
• Human-in-the-Loop validation to ensure regulatory compliance.

Regulatory-Grade Data Curation and Quality Control

• Custom review workflows with role-based access control.
• Version tracking and change history for full auditability.
• Quality metrics monitoring and team output dashboards.
• Advanced analytics and guidelines for annotation consistency.
• Comments and feedback mechanisms within projects.
• Full audit trails to support compliance requirements.

Enterprise Security Features

• Role-based access control (RBAC) for managing user permissions.
• Multi-factor authentication (MFA) and Active Directory/LDAP integration.
• Support for cloud, on-premise, and air-gapped deployments.
• Robust encryption and real-time event monitoring dashboards.
• Tamper-proof audit logs and identity provider integration.
• Zero data sharing with John Snow Labs or third parties when deployed on-premise or in air-gapped environments.

Scalability and Infrastructure

• Kubernetes-based auto-scaling architecture that adjusts computational resources based on demand.
• Supports concurrent multi-user access and high document throughput.
• GPU support for faster pre-annotation and model training.
• No limitations on the number of projects, users, documents, models, prompts, rules, annotations, pre-annotations, or training runs.
• Minimum hardware requirement for on-premise deployment: 8-core CPU, 32 GB RAM, 512 GB SSD; recommended for model training: 16-core CPU, 64 GB RAM, 512 GB SSD.

Licensing and Pricing

• Available as a pay-as-you-go subscription on AWS Marketplace and Azure Marketplace.
• On-premise deployments require a license key, available by contacting John Snow Labs support.
• Pay-as-you-go billing charges only for active feature usage and duration.
• Enterprise on-premise pricing options are available for teams with continuous annotation workflows.
• The software subscription includes GPU support, Visual Document Understanding, LLM and zero-shot prompt pre-annotation, access to 2,600+ healthcare models, and premium support.

Generative AI Lab can be deployed on AWS, Azure, or on-premise Kubernetes and single-machine environments, making it suitable for high-compliance industries including healthcare, life sciences, finance, and insurance. Onboarding resources include a QuickStart guide, video tutorials, and step-by-step documentation. John Snow Labs also offers training, certification programs, and live workshops covering text annotation, Spark NLP, and healthcare-specific NLP.