About the job
About Granica
Granica is an innovative AI research and infrastructure firm dedicated to creating reliable, steerable representations of enterprise data.
We establish trust through Crunch, a policy-driven health layer optimizing large tabular datasets for efficiency, reliability, and reversibility. Utilizing this foundation, we are developing Large Tabular Models—systems designed to learn cross-column and relational structures, delivering trustworthy answers and automation with integrated provenance and governance.
Our Mission
Current AI capabilities are hindered not only by model design but also by the inefficiencies of the data that supports it. At scale, each redundant byte, poorly organized dataset, and inefficient data pathway contributes to significant costs, latency, and energy waste.
Granica’s mission is to eliminate these inefficiencies. We leverage groundbreaking research in information theory, probabilistic modeling, and distributed systems to craft self-optimizing data infrastructure: systems that continually enhance how information is represented and utilized by AI.
Led by Prof. Andrea Montanari from Stanford, Granica’s Research group merges advances in information theory with learning efficiency in large-scale distributed systems. We collectively believe that the next significant leap in AI will originate from innovations in efficient systems, rather than merely larger models.
Granica is at the forefront of developing a new category of structured AI models: foundational models designed to learn and reason from the relational, tabular, and structured data that drives the global economy. While many focus on unstructured text or media, we are venturing into the next frontier: systems capable of comprehending and reasoning over structured information.
Your Contributions
Create and prototype algorithms that form the core of structured AI, enhancing representation learning and efficient information modeling for enterprise and tabular data at petabyte scale.
Develop adaptive learners merging statistical learning theory with systems optimization at scale, contributing to a new generation of foundational models for structured information.
Design architectures that unify symbolic, relational, and neural components, enabling AI systems to reason directly over structured enterprise data.
Construct cost models and optimization frameworks that enhance the efficiency of structured learning, both computationally and economically.

