An independent AI research lab for problems that do not fit one institutional box.
studio1804 is an independent AI research lab. Our work centers on systems that handle real-world complexity in language, archives, evidence, and the infrastructure that makes them trustworthy. Some projects are research; some become public-knowledge institutions; some are commercial; the structure follows the work.
Rasin began at studio1804 as a way to help researchers, students, journalists, and Haitian diaspora communities explore Haiti and its diaspora through original sources. As it moves toward public-knowledge infrastructure for scholars, archives, and Haitian communities, an independent nonprofit structure is being planned, with studio1804 expected to serve as the founding technical partner.
The research question behind Rasin is practical: how can archives with low-resource languages become searchable, citable, and technically durable without depending on black-box systems that can't be inspected, evaluated, or handed off? Haitian Kreyòl has no standard retrieval benchmark; Rasin's internal golden set is a starting point. The archives exist, scattered across three continents, and no system has connected them as a single research surface.
That same research posture extends beyond Rasin: multilingual retrieval, citation verification, portable infrastructure, and the security of autonomous AI systems at the agent-tool boundary. Where the work becomes reusable, we publish methods and evaluation so others can inspect, test, and adapt them.
The lab continues to hold varied work, including AI infrastructure security and other projects in formation. If you are a scholar, librarian, archivist, program officer, or partner working where language and evidence matter, we'd like to hear from you.
Language Systems
We build for multilingual records, low-resource languages, spelling variation, and archives where the language of the question may not match the language of the source.
Evidence Systems
Retrieval, citation, and claim-provenance workflows keep answers tied to original documents, pages, collections, and the uncertainty around them.
Infrastructure
Our work includes portable archive infrastructure and security research for AI systems that invoke external tools, services, and data sources.
Project Formation
Some projects become public-knowledge institutions, some remain lab research, and some may become commercial work. The structure follows the problem.
Searchable
Historical collections should be discoverable across language, spelling variation, and archival silos.
Citable
Answers should point back to exact documents and pages, not collapse evidence into unsupported summaries.
Exportable
Partners should retain usable outputs: OCR text, metadata, indexes, embeddings, and citation records.
Reproducible
Methods should be documented well enough for researchers and archivists to inspect, test, and extend.
Multilingual
Haitian Kreyòl, French, English, and Spanish are handled as part of one research surface.
Auditable
The system should make its sources, retrieval paths, and limits visible to the people evaluating it.
Ricardo founded the lab to build AI systems that work across languages, stay grounded in evidence, and can be operated by the institutions that rely on them.
His work spans Rasin, multilingual retrieval and citation systems for cultural collections, and research on securing autonomous AI infrastructure at the agent-tool boundary.
Before founding the lab, Ricardo built production systems and advanced prototypes for enterprise and federal clients. That experience shapes the lab's product discipline: observability, reliability, maintainability, evaluation, and handoff from the start.
Rasin is being developed through invited consultation.
The scholarly and institutional network around Rasin is being formed through invited conversations with scholars in Haitian history, Kreyòl computational linguistics, and digital humanities; with U.S. academic libraries that hold relevant collections; and with Haitian institutions whose work we index. Formal advisory roles will be named as commitments are confirmed. If you teach, research, or steward materials in any of these areas, we'd like to hear from you.
Haitian history, Kreyòl computational linguistics, archival science
Caribbean collections, Haitian diaspora archives
libraries, universities, civil-society foundations
Work With Us
Funders, scholars, and archives each need a different doorway into the work. Start with the path that fits your role.