PhD Intern - Computing Cost Modeling and Optimization

** Overview*
* At PNNL, our core capabilities are divided among major departments that we refer to as Directorates within the Lab, focused on a specific area of scientific research or other function, with its own leadership team and dedicated budget.

Our Science & Technology directorates include National Security, Earth and Biological Sciences, Physical and Computational Sciences, and Energy and Environment. In addition, we have an Environmental Molecular Sciences Laboratory, a Department of Energy, Office of Science user facility housed on the PNNL campus.

The Physical and Computational Sciences Directorate's (PCSD's) strengths in experimental, computational, and theoretical chemistry and materials science, together with our advanced computing, applied mathematics and data science capabilities, are central to the discovery mission we embrace  our most important resource is our people-experts across the range of scientific disciplines who team together to take on the biggest scientific challenges of our time. 

The Advanced Computing, Mathematics, and Data Division (ACMDD) focuses on basic and applied computing research encompassing artificial intelligence, applied mathematics, computing technologies, and data and computational engineering. Our scientists and engineers apply end-to-end co-design principles to advance future energy-efficient computing systems and design the next generation of algorithms to analyze, model, understand, and control the behavior of complex systems in science, energy, and national security.

** Responsibilities*
* Realizing the potential for science to leverage on-demand cloud scaling and resource diversity requires the ability to control and manage costs. Scientists must have the ability to request SLO constraints with respect to a job's cost. Job orchestrators must be able to schedule task parallelism, manage data objects, select compute instances, and assign storage resources while also tracking costs and ensuring they stay within budgets.

Although cloud vendors provide cost calculators, they provide no ability to specify cost within a SLO constraint for specific jobs. The fundamental problems are that billing data is delayed, many services bill asynchronously, jobs may create downstream resources, charges for shared resources (network, storage, logging) accrue separately, and the ability to specify resource limits is constrained only to particular and local services.

It is especially difficult to estimate charges for a distributed set of resources or for agentic workflows that generate dynamic or unpredictable tasks.

PNNL's Future Computing Technologies group seeks an accomplished PhD Intern to explore methods for the characterization and modeling workflow resource usage and cost accumulation within the cloud. Relevant research topics include:

+ Developing job resource telemetry, cost introspection, modeling, and prediction, to reason about expected vs. actual SLO cost.

Cross-platform job control mechanisms that enable appropriate alerts as the job progresses, soft landings through checkpointing, and hard stops if necessary.

+ Optimized job execution policies adapted to and reinforced by the cost profile and reasoning

The successful applicant will work within the Future Computing Technologies group and have demonstrated expertise in a topic closely related to performance modeling and scientific workloads. The researcher should be creative, self-motivated, and ready to publish at top-tier venues.

** Qualifications*
* Minimum Qualifications:

+ Candidates must be currently enrolled/matriculated in a PhD program at an accredited college.

+ Minimum GPA of 3.0 is required.

Preferred Qualifications:

+ Pursuing a degree in computer science, data science, or related field.

+ Familiar with topics such as distributed and continuum computing, vector databases, performance modeling, storage and memory systems, etc.

** Hazardous Working Conditions/Environment*
* Not Applicable

** Testing Designated Position*
* This is not a Testing Designated Position (TDP).

** About PNNL*
* Pacific Northwest National Laboratory (PNNL) is a world-class research institution powered by a highly educated, diverse…