Research Engineer - Applied AI

Granica is an AI research and systems company building the infrastructure for a new kind of intelligence: one that is structured, efficient, and deeply integrated with data.

Our systems operate at exabyte scale
, processing petabytes of data each day for some of the world’s most prominent enterprises in finance, technology, and industry. These systems are already making a measurable difference in how global organizations use data to deploy AI safely and efficiently.

We believe that the next generation of enterprise AI will not come from larger models but from more efficient data systems
. By advancing the frontier of how data is represented, stored, and transformed, we aim to make large-scale intelligence creation sustainable and adaptive.

Our long-term vision is Efficient Intelligence
: AI that learns using fewer resources, generalizes from less data, and reasons through structure rather than scale. To reach that, we are first building the Foundational Data Systems that make structured AI possible.

The Applied AI Research Team sits at the center of this mission. Your work will take the ideas emerging from fundamental research and turn them into practical algorithms, optimized pipelines, and production-ready systems that operate across petabytes of structured enterprise data.

This is a high-ownership role for engineers who can think like researchers and build like systems experts. You will translate theory into measurable performance improvements and help define the foundations of structured AI.

Turn research into real systems

• Transform foundational ideas from Granica Research and Prof. Andrea Montanari’s group into scalable algorithms and experimental prototypes.

• Build the evaluation harnesses, metrics, and datasets that reveal real signal from research concepts.

• Define and refine the metrics that determine progress in structured AI.

Invent and optimize algorithms for structured AI

• Develop efficient learning methods for relational, tabular, graph, and enterprise data.

• Prototype representation learning architectures and compression-aware models for large-scale structured information.

Build high-performance learning pipelines

• Implement fast training and inference loops using PyTorch, JAX, or custom kernels.

• Optimize memory, compute, and data-movement paths with a focus on cost, latency, and throughput.

Integrate symbolic, relational, and neural components

• Design hybrid learning systems that reason over structured data natively, not through text intermediaries.

Collaborate deeply across teams

• Work with Research Scientists to validate hypotheses at scale.

• Work with Systems Engineers to integrate your algorithms into Granica’s core data platform.

• Work with Product Engineering to ship features that power live enterprise workloads.

Iterate fast and measure everything

• Run controlled experiments, analyze performance deltas, and deliver results with clear benchmarks.

• Drive the loop from prototype to production, improving the system each cycle.

Technical Depth

• Strong background in machine learning, probabilistic modeling, optimization, or distributed learning.

• Experience building and tuning algorithms for structured, tabular, graph, or relational data.

• Ability to reason from first principles about efficiency, scaling behavior, and information flow.

Systems Ability

• Hands‑on experience with PyTorch, JAX, Tensor Flow, or custom ML kernels.

• Strong Python skills and familiarity with systems languages such as Rust, C++, or CUDA.

• Experience with large‑scale data pipelines, model evaluation frameworks, or distributed systems.

Applied Mindset

• Demonstrated ability to turn theoretical concepts into performant, reliable code.

• Comfort working in ambiguous research environments while delivering measurable outcomes.

• Curiosity for how structure and efficiency reshape the next generation of AI.

Bonus Experience

• Structured representation learning, tabular or multimodal models, or relational ML.

• Distributed data systems, query engines, or vector/embedding infrastructure.

• Open‑source contributions or collaborative research bridging theory and production.

The world’s most valuable data is structured
.…