AI Engineer
← All organizations

Enterprise AI agents and retrieval-augmented generation

Contextual AI

Contextual AI builds a platform for enterprise AI agents that retrieve information from private company data, cite sources, and execute tasks. Agent Composer lets developers combine search, reranking, tool calls, and generation through visual or YAML workflows. Engineering, support, research, and financial-services teams can use these agents to investigate logs, answer technical questions, and extract structured information. The platform connects to repositories such as SharePoint and Google Drive and is accessible through a UI, APIs, and SDKs.

Founded in 2023 by Douwe Kiela and Amanpreet Singh, the company developed RAG 2.0, which underpins its Contextual Language Models by jointly optimizing the retriever and language model as an integrated system rather than assembling frozen components. The work extends Kiela’s earlier research at Facebook AI Research, where his team introduced retrieval-augmented generation in 2020.

Contextual AI charges by usage and offers managed-cloud, private-cloud, and customer-VPC deployment; its listed customers include Qualcomm and HSBC. In May 2026, Google DeepMind agreed to hire more than 20 researchers, including Kiela, in a licensing deal covering Contextual’s technology—not an acquisition of the company. The remaining product team was to remain under interim CEO Jay Chen.

contextual.ai ↗

2 talks

Newest first

3 speakers at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Nina LopatinaStaff developer advocateRajiv ShahAgentic AI engineerDouwe KielaGoogle DeepMind researcher

Start here

  1. Specialized RAG Agents: Lessons learned from deploying complex AI systems in production

    Start with the essay based on Douwe Kiela's talk to understand why the complete retrieval system matters more than model capability alone, and how engineering effort connects to business value.

    Douwe KielaAI Engineer Summit 2025

  2. Forget RAG Pipelines—Build Production-Ready AI Agents in 15 Minutes

    Watch the workshop for notebook, API, and graphical-interface approaches to creating managed RAG agents, plus questions about JavaScript integration and data residency.

    Nina Lopatina · Rajiv ShahAI Engineer World's Fair 2025

Messages from the stage

Plan beyond the pilot

Kiela describes how larger document collections, more users, and varied use cases complicate production readiness. He also argues that teams must understand and manage remaining errors after reaching an acceptable accuracy threshold.

Evaluate agents with natural language

Lopatina and Shah’s workshop demonstrates LMUnit alongside financial and spurious-correlation examples, making evaluation a concrete part of the agent-building workflow.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-28