Computer Science at Orrick Research

Research in efficient AI inference, scientific computing, optimization, cybersecurity, and reproducible high-performance computing.Orrick Research develops computational methods aimed at increasing performance, reducing energy use, and making difficult problems more practical to solve.

Our work spans artificial intelligence, scientific computing, optimization, and cybersecurity. We study how mathematical insight, algorithm design, and software execution can work together to improve what existing computing systems can accomplish.

Led by Brian Orrick, this research connects decades of technical experience with an active program of invention and product development.

Running large language models requires substantial computation, memory movement, and energy. These demands affect response time, operating cost, and the scale at which AI can be deployed.

Our AI research investigates ways to make model execution more efficient, with particular attention to inference performance and the use of existing hardware.

The objective is to deliver more useful output for the resources consumed. This includes examining the relationship among throughput, latency, memory requirements, energy use, and output quality.

Through Orrick Industries, this work also includes MLPerf Inference submissions, connecting research developments with standardized performance evaluation.

Computational capacity shapes the questions scientists can investigate. Faster calculations can make it possible to explore more parameter combinations, compare additional models, and examine a system at greater resolution.

Orrick Research develops approaches to accelerating demanding scientific workloads. Our investigations connect mathematical reformulation with algorithm development and implementation, examining where work can be reduced, reused, or organized more effectively.

Our mathematics and physics programs provide demanding applications for this research, including dynamical systems, gravitational calculations, and the numerical evaluation of theoretical models.

The goal is to make computational methods more useful for discovery while preserving the accuracy required by the underlying scientific question.

Optimization problems arise whenever a system must select an effective solution from many possible choices.

We investigate methods for identifying and using the structure within these problems. Research considers how constraints interact, how candidate solutions can be evaluated, and how search can be organized to reduce unnecessary work.

Potential applications include planning, scheduling, resource allocation, and scientific model development.

Evaluation considers solution quality alongside runtime, memory consumption, and scaling behavior. Each method must be understood in terms of the problems it addresses and the conditions under which it performs well.

Effective cybersecurity depends on both sound foundations and practical execution.

Our cybersecurity research draws on experience in computing systems, mathematical methods, and technical operations. We investigate approaches to protecting information and system integrity while accounting for the performance and integration requirements of real deployments.

This work connects security research with the broader computing program: understanding how systems behave, identifying weaknesses in their assumptions, and developing methods that can be examined and tested.

A computational improvement becomes meaningful when its benefits can be demonstrated under clearly defined conditions.

Our evaluation priorities include:

• Throughput: how much useful work a system completes.
• Latency: how quickly individual requests receive results.
• Energy efficiency: how much energy is required for a defined workload.
• Accuracy and output quality: whether the result meets the task’s requirements.
• Resource use: the memory and hardware needed to achieve the result.
• Reproducibility: whether a comparison can be repeated under the stated conditions.

These measures help connect research progress with the requirements of scientists, developers, and organizations deploying the technology.

Orrick Research & Development is the research arm of Orrick Industries.

The relationship connects mathematical and computational investigations with the engineering needed to develop practical technology. Promising methods can progress from an initial formulation through prototypes, performance evaluation, and product development.

Our objective is to turn computational insight into tools that deliver measurable value: faster results, lower resource requirements, and greater capacity to address demanding problems.

We welcome conversations with researchers, engineering teams, and organizations working on AI infrastructure, scientific computing, optimization, and cybersecurity.

For a technical inquiry, describe the workload or research question, the current constraints, and the outcome you want to improve. This provides a useful starting point for discussing research collaboration, evaluation, or commercial development.