FORMAL SEMANTICS

Dr. Renée J. Miller

Canada Excellence Research Chair in Data Intelligence

Cheriton School of Computer Science, University of Waterloo

Renée Miller has spent decades on how to integrate, exchange, and make trustworthy sense of large volumes of complex, heterogeneous data — and turned it into practical systems the field actually uses. Her group is giving UIR’s tool-call witness principle a formal backbone in provenance and schema-mapping theory.

Fellow, Royal Society of Canada

PECASE recipient

NSF CAREER Award

Foundational work on provenance semirings & data exchange

the partnership

Formal rigor, meet algebraic infrastructure.

Every agent framework today asks you to trust that a tool call did what it says it did. UIR — our algebraic grammar for portable agent intent — was built to end that trust exercise. Its core bet: a tool call doesn’t just perform an intent, it witnesses it, and that witness is part of the intent’s identity, not a log line bolted on after the fact.

That bet needs more than good design. It needs formal grounding in data integration, schema mapping, and provenance — the ground Dr. Miller and her group have been standing on. This collaboration brings UIR’s next chapter into that foundation.

the research

Three potential directions, one thesis at a time.

The collaboration is structured as co-supervised PhD research — Algedonic providing the systems and roadmap context, Dr. Miller’s group providing the formal rigor and publication path.

DIRECTION 01

Provenance Semirings for Agent Intent

Giving UIR’s tool-call witness principle a real mathematical backbone, extending provenance semiring theory from queries to agent action traces.

DIRECTION 02

Intent Exchange as Data Exchange

Treating the translation of intent across MCP, A2A, and ACP as a schema mapping problem, with the correctness guarantees data exchange theory demands.

DIRECTION 03

Capability Discovery over Tool Lakes

Adapting data lake discovery so agents can find the right tool for a given intent — out of thousands — at scale.

why it matters

“Every agent framework today asks you to take a tool call’s word for what it did. Provenance, schema mapping, and capability discovery aren’t three separate favors we’re asking of academia — they’re the same open problem in three different disguises, one the database field has already spent decades rigorously solving. So instead of inventing our own ad hoc notion of proof, we partnered with someone who helped define what proof actually means for data, and is now helping us define what it means for intent.”

— Sandeep Gopisetty, Founder & CEO, Algedonic

← Back to research portfolio