Enterprise software, hybrid cloud, AI and consulting
IBM
IBM provides software, consulting and computing infrastructure that help enterprises integrate AI into business systems and workflows. Its portfolio includes watsonx, Red Hat, IBM Z mainframes, servers and storage. Organizations use these offerings to build AI applications, modernize existing systems and connect operations across hybrid cloud environments. Confluent streams live operational events into watsonx.data, supplying models and agents with continuously updated enterprise data, lineage and policy controls.
Charles Ranlett Flint organized the 1911 merger that created Computing-Tabulating-Recording Company. Thomas J. Watson Sr. joined in 1914 and renamed it International Business Machines in 1924; today, Arvind Krishna is chairman, president and CEO. IBM’s research contributions include Edgar F. Codd’s 1970 relational model, which let users query related tables without knowing their physical storage layout. Subsequent System R work developed SQL and cost-based query optimization.
IBM completed its Confluent acquisition in March 2026 at approximately $11 billion in enterprise value. At closing, IBM said more than 6,500 enterprises relied on Confluent, giving its data-streaming business an established customer base. IBM also completed its acquisition of HRL Laboratories in August 2026, adding silicon-spin qubit and quantum-sensing expertise alongside its superconducting quantum-computing research. HRL brings capabilities in materials, cryogenics and control electronics; financial terms were undisclosed.
3 talks
Newest first
3 speakers at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here
- Harnesses in AI: A Deep Dive
Start with the essay based on Tejas Kumar's talk to follow a browser-agent experiment that isolates the effect of harness improvements from changes to prompting.
Tejas KumarAI Engineer Europe 2026
- Tool Calling Is Not Just Plumbing for AI Agents
Watch Roy Derks’s talk for a comparison of traditional and meta tool calling and how external APIs, databases, search, and computation extend language models.
Roy DerksAI Engineer Summit 2025
- OpenRAG: An open-source stack for RAG — Phil Nash
Watch Phil Nash’s talk to learn about configurable embeddings, offline deployment, and application customization in OpenRAG.
Phil NashAI Engineer Europe 2026
Messages from the stage
Control the runtime around the model
Kumar’s demonstration keeps prompts unchanged while adding runtime controls. Iteration limits, context compression, and deterministic login handling make the surrounding system responsible for behavior the model cannot reliably manage alone.
Treat tools as engineered interfaces
Derks emphasizes deliberate tool descriptions, typed arguments, and structured outputs. He also discusses framework-independent tool platforms and generating tools dynamically from GraphQL or SQL schemas.
Retrieval starts with document processing
Nash presents OpenRAG as a combination of Docling processing, OpenSearch retrieval, and Langflow orchestration. His ingestion examples include OCR, speech recognition, and hierarchical chunking, alongside an explanation of why larger context windows do not eliminate retrieval challenges.
Affiliations reflect each recorded session, not necessarily current employment.