IBM watsonx, plus the ecosystem that makes it work.

High level

An end-to-end studio for AI builders to develop generative and traditional ML applications using IBM Granite, Llama, Mistral, and other foundation models — all behind your firewall or in the cloud of your choice.

• GenAI apps
• Fine-tuning
• RAG
• Hybrid deployment

Deep dive

• Prompt Lab and Tuning Studio for prompt engineering, fine-tuning, and parameter-efficient tuning (LoRA / QLoRA).
• Model library spanning IBM Granite, Meta Llama, Mistral, and a curated set of open and third-party models — plus bring-your-own model support.
• Built-in RAG patterns, vector indexing, and integration with watsonx.data for grounding on enterprise content.
• Deployment to hybrid environments — IBM Cloud, AWS, Azure, on-prem, or Red Hat OpenShift — with the same artifacts.
• Native hooks into watsonx.governance for evaluation, drift detection, and audit trails.

High level

Build, deploy, and manage AI agents that automate work across HR, sales, procurement, and IT — with prebuilt skills for the SaaS apps your business already runs on.

• Digital workers
• Workflow automation
• HR / IT / Sales ops

Deep dive

• Agent Builder for low-code authoring of agents, tools, and multi-step workflows.
• 1,000+ prebuilt skills connecting to Salesforce, Workday, SAP, ServiceNow, Microsoft 365, and more.
• Agentic orchestration — agents can call other agents, handle long-running tasks, and stay in the loop with humans for approvals.
• Conversational interface that lives inside Slack, Teams, or your own app via API.
• Enterprise governance: role-based access, audit logging, and grounding in your own data and policies.

High level

Formerly IBM Knowledge Catalog and Manta lineage, watsonx.data intelligence is the active metadata layer that helps teams discover, trust, and govern data and AI assets across the stack.

• Data catalog
• Lineage
• Data quality
• AI readiness

Deep dive

• Automated data discovery and classification across cloud, on-prem, and SaaS sources.
• End-to-end column-level lineage — from source systems through pipelines, BI dashboards, and AI features.
• Data quality scoring, rule-based validation, and remediation workflows.
• Business glossary and semantic enrichment so non-technical users can find and trust data.
• Tight coupling with watsonx.data and watsonx.governance for AI-ready, policy-aware data.

High level

A single platform for batch ETL/ELT, change data capture, real-time streaming, and data replication — built on the assets of DataStage, StreamSets, and IBM's data fabric portfolio.

• ETL / ELT
• CDC
• Streaming
• Data fabric

Deep dive

• Visual pipeline designer plus code-first mode (Python, SQL) for hybrid teams.
• Hundreds of connectors across cloud warehouses, lakes, mainframes, and SaaS — with native push-down to Snowflake, Databricks, and watsonx.data.
• Change data capture and stream processing for real-time pipelines (StreamSets DataOps lineage included).
• DataOps features: drift detection, pipeline versioning, and observability across runs.
• One control plane for batch + streaming + replication — reducing tool sprawl.

High level

A toolkit for AI risk, compliance, and lifecycle governance — covering both generative and predictive models, regardless of where they were built.

• AI risk
• Compliance
• Model monitoring
• EU AI Act

Deep dive

• Model inventory across watsonx.ai, SageMaker, Vertex, Azure OpenAI, and custom models.
• Automated AI factsheets — data sources, training, evaluation, approvals — for audit and regulator-ready documentation.
• Continuous monitoring for drift, bias, toxicity, hallucination, and performance.
• Policy-as-code mapped to EU AI Act, NIST AI RMF, and internal frameworks.
• Workflow integration with risk, compliance, and model-risk-management teams.

High level

Confluent Cloud and Platform deliver a complete data-streaming platform built around Apache Kafka and Flink — the real-time substrate that feeds watsonx.data, watsonx.ai, and event-driven apps.

• Kafka
• Flink
• Real-time AI
• Event-driven apps

Deep dive

• Fully managed Kafka with 99.99% SLA, elastic scaling, and infinite storage tiers.
• Stream Governance — schema registry, stream lineage, and data quality rules across topics.
• Apache Flink for stateful stream processing in SQL, Java, or Python.
• 120+ pre-built connectors (CDC, SaaS, cloud) to land and move data without custom code.
• Direct integration paths into watsonx.data (Iceberg) and watsonx.ai for real-time RAG and feature pipelines.

High level

Unstructured.io transforms PDFs, slides, images, emails, HTML, and 25+ file types into clean, chunked, embedding-ready data — the missing front-end for any RAG pipeline on watsonx.ai.

• RAG ingestion
• Document parsing
• Chunking
• Connectors

Deep dive

• High-fidelity document parsing that preserves tables, layout, headings, and figures.
• Smart chunking strategies (by title, semantic, custom) tuned for retrieval quality.
• Native connectors for SharePoint, Google Drive, S3, Confluence, Notion, Salesforce, and more.
• Enterprise platform with VPC / on-prem deployment for sensitive content.
• Output formats ready for watsonx.data vector tables, Milvus, and any embedding model in watsonx.ai.

High level

Bob is IBM's AI-powered development partner that works alongside engineers in the IDE and the terminal — built for enterprise codebases where modernization, security, and compliance matter.

• Agentic coding
• Legacy modernization
• Code review
• Enterprise guardrails

Deep dive

• Agentic Modes — Bob takes on a defined role (architect, modernizer, reviewer, etc.) with the right tools for the job; teams can author custom modes for their stack.
• Literate Coding — describe intent in natural language and Bob generates the implementation in-context, no chat-window context-switching.
• Real-time code reviews catch complexity and refactoring opportunities as you type, fixable inline or triaged in the Bob Findings panel.
• Bob Shell extends the same capabilities into the terminal — local dev through production pipelines.
• Native access to the IBM enterprise ecosystem: HashiCorp, Red Hat, Instana, and more, directly from the IDE.
• Built-in guardrails — approval-based modes and refusal to hallucinate on unknown APIs (e.g., RPG op-codes), important for regulated environments (HIPAA, FedRAMP).