AI Context Engineering for R&D Knowledge Using a Multi-Model Database
summary
A reproducible pipeline ingests heterogeneous R&D content into MarkLogic multimodel, extracts metadata and chemical templates, adds BM25 and vector indexes, then runs hybrid retrieval plus a knowledge-graph layer that selects from ~36 domain tools for orchestration. An agent uses target normalization, auto-generates Optic API queries, pulls reactions into a table, and returns three viable synthesis routes while flagging toxicity and regulatory risks; iterative feedback fixes failed code generations so the proof-of-concept pipeline actually works as an end-to-end graph RAG system.
AI-written summary · about
chapters (8)
- 0:00Introduction and journey
- 4:43Combining MarkLogic and GenAI
- 6:30Context engineering fundamentals
- 6:51Retrieval augmented generation
- 12:53AI agents and R&D
- 18:08Demo walkthrough
- 29:32Future of AI in R&D
- 31:37Q&A session
related talks
15mGraphRAG-Powered AI Agents: Transforming Corporate Metadata into Intelligent Knowledge
33mSmarter Industrial Fault Diagnosis with Knowledge Graphs and Large Language Models
44mBuilding Knowledge Graphs- 44mBuilding Knowledge Graphs
36mVibing with Data -- Multi-Agent Data Modeling and Constructi...
01h 00mFrom Planning To Production-Ready RAG
31mBuild Intelligent AI Agents With Context Engineering
01h 06mDesign and Implementation of the RelationalAI Knowledge Graph Management System