The Challenge
The Observatorio de Arquitectura Latinoamericana Contemporánea (ODALC) is a research center dedicated to studying contemporary architecture in Latin America. Over decades of scholarship, however, the field’s most influential work became scattered across hundreds of journals, countries, and languages. More than 800 foundational articles existed, but they were difficult to find, compare, and study together.
As a result, researchers and students relied on manual reading, disconnected archives, and individual expertise — making synthesis slow, limiting scale, and restricting access to knowledge beyond select institutions. This fragmentation also posed deeper risks: It constrained epistemic diversity, reinforced paywalled access, and made it harder to study Latin American architectural theory as a coherent, evolving body of thought.
The Vision
ODALC set out to build a shared, responsible AI foundation capable of organizing and analyzing Latin American architectural theory at scale, without sacrificing cultural context, epistemic diversity, or academic integrity. The goal was to create governed, AI-augmented workflows that faculty, students, and researchers could all use, reflecting a challenge many enterprises face when serving users with different levels of expertise and domain authority.
By bringing fragmented knowledge into a shared, self-service foundation, ODALC opened access to collective intelligence that was previously siloed — enabling more people to search, compare, and build on the same body of research while preserving trust, context, and accountability.
The Execution
Using Dataiku, The Universal AI Platform™, ODALC built an agentic research system that transforms fragmented academic archives into a unified, interactive knowledge platform.
At the core of the solution is a Dataiku-powered knowledge agent using retrieval-augmented generation (RAG), enabling users to query more than 800 multilingual academic articles in natural language and quickly surface relevant theories, authors, and conceptual frameworks.
Behind the scenes, Dataiku orchestrates NLP pipelines that ingest, clean, and analyze unstructured texts at scale. These workflows extract keywords, cluster concepts, map co-occurrence patterns, and track the evolution of architectural theories over time — creating a standardized and reusable research foundation.
Built with Dataiku’s no- and low-code capabilities, the platform is accessible to non-technical users and functions as a shared, self-service research agent. As a result, ODALC reduced manual processing time by 80% while establishing a scalable, governed model for AI-augmented humanities research.
