Job search
DOE CSGF Application Guide: Choose the Math/CS or Science Track and Plan the Lab Practicum
Compare research-track fit, build an advisor-ready course plan and track application materials while keeping the laboratory practicum distinct from employment.
Choose a DOE Computational Science Graduate Fellowship track by the research contribution you intend to make. Broadly reusable methods and technologies that improve high-performance computing point toward Math/CS; computational research centered on a specific science or engineering problem points toward Science & Engineering. Both tracks require an interdisciplinary course plan and a laboratory practicum outside your thesis research.
Checked October 23, 2026: the current application is for the 2027–2028 incoming cohort, due January 7, 2027. This guide reviews public program documents and uses two invented academic-planning cases. It does not assess a real applicant’s eligibility or describe an application we submitted. LandOffer publishes this guide; if your immediate goal is employment rather than doctoral study, browse recent roles on LandOffer.ai as a separate route.

Start with the research question, not the department label
The track descriptions distinguish enabling HPC technologies from the use of HPC to investigate a particular scientific or engineering problem. Both involve computational science. The choice is about the proposed research, rather than whether you happen to write Python or belong to a department with “computing” in its name.
For Math/CS, the detailed track page requires doctoral study in applied mathematics, statistics, computer science, computer engineering, computational science, or an academic equivalent. It describes broadly relevant enabling work such as compilers, scalable solvers, programming models, and performance tools. In every case, the research must support more effective use of emerging high-performance systems. A general software interest without that connection is not enough to explain track fit.
The page also distinguishes quantum algorithms or software from using scientific computing to design or control quantum information systems. A departmental exception can allow qualifying algorithms/software research in a science or engineering field. “Quantum” alone therefore does not resolve the choice. Frame the actual question and contribution, then ask the program about an unresolved boundary rather than selecting a track because its name sounds more impressive.
Check the program rules separately from research fit
The eligibility requirements specify U.S. citizenship or lawful permanent residence and full-time, uninterrupted Ph.D. study at an accredited U.S. university. They list eligible academic-stage categories at application, including undergraduate seniors and first-year graduate students, alongside conditions for current master’s students and other entrants. They exclude second-year-or-beyond graduate students and concurrent professional doctoral degrees.
Do not reduce those categories to “any STEM graduate student.” Prior graduate study and institution changes have detailed rules. Use the current portal FAQ and the program’s own response for an ambiguous record; this article cannot decide individual eligibility from a label such as “first year.” Research fit, formal eligibility, university admission, and fellowship selection are separate questions.
The benefits page currently lists a $45,000 yearly stipend, full tuition and required fees during the appointment, a $1,000 annual professional-development allowance, and up to four years of support subject to renewal. These are fellowship benefits, not a DOE employee salary quote. The practicum and professional-development allowances are different provisions; do not combine them into an invented relocation budget.
The program also limits required departmental teaching and requires advance approval for additional graduate responsibilities. Discuss those conditions with the department before assuming the fellowship can be combined with an existing assistantship. Funding supports an academic commitment, with conditions on additional graduate responsibilities; it does not promise a laboratory job.
Build the course plan before polishing the application
The Program of Study requirements are the same for both tracks. They cover science/engineering applications, mathematics/statistics, computer science/computer engineering, and an introductory HPC course. The published requirement is two semester courses or three quarter courses in each disciplinary area, plus one HPC course.
Math/CS does not remove the science requirement. Its track page says courses from CS or computer-engineering departments belong under that category, not Science & Engineering Applications. Read the actual course content and current application guidelines with an advisor; do not rename a CS course to fill a missing science slot.
Timing depends on entry stage. Someone beginning doctoral study with the fellowship must complete the POS before the start of the fourth fellowship year and at least one course in each disciplinary area within the first two years. Someone beginning the fellowship in the second graduate year must finish before the third fellowship year. Proposed later changes require advance review and approval. A plausible timetable exposes missing prerequisites and overloaded terms. It is preparation, not proof that courses or substitutions have been accepted.
Worked Math/CS plan: reusable scheduling methods
The following people, research plans, courses, institutions, dates of personal actions, and application records are invented. They are planning examples, not actual applicants or eligibility determinations.
Iona Mercer is preparing a hypothetical doctoral research direction in computer science: scheduling irregular tasks across heterogeneous HPC resources. Her intended contribution is a broadly reusable scheduling method. She would evaluate it across several computational workloads instead of making one physical phenomenon her central research question.
Completed track rationale: “My proposed question is how to assign irregular tasks across heterogeneous HPC resources while accounting for communication overhead. The contribution would be a reusable scheduling method and an evaluation of its limits across workload families. A scientific workload would help evaluate the method, but the research objective is the method’s general behavior.” This explains a Math/CS research direction without claiming a working algorithm or measured improvement.
Completed draft semester POS for advisor discussion:
| Area | Proposed courses and timing | Purpose |
|---|---|---|
| Science & Engineering | Continuum Mechanics, Fall 2027; Computational Materials, Spring 2028 | Understand scientific workloads beyond CS |
| Mathematics & Statistics | Numerical Linear Algebra, Fall 2027; Probability for Computation, Fall 2028 | Analyze algorithms and uncertainty |
| CS & Computer Engineering | Parallel Algorithms, Spring 2028; Compiler Construction, Spring 2029 | Study scheduling and execution behavior |
| HPC | Introduction to High-Performance Computing, Spring 2028 | Establish the required HPC foundation |
This fictional seven-course plan puts at least one course from each of the three disciplinary areas in the first two years and finishes before a fourth fellowship year. It assumes doctoral study and funding begin together; a different entry stage needs a different timing check. The course names are not claims that a real university offers them or that the committee would approve their classification.
Completed application preparation record: Iona’s research description is a proposal, with no invented benchmark result. Her POS is marked “draft for advisor discussion.” She has identified three hypothetical referees by relationship—research supervisor, advanced-course instructor, and project mentor—without fabricating their letters. Official transcript requests would come from each attended institution. In this invented status record, two letters are received, one is pending, and the application is still a draft. It is not marked complete merely because all essay boxes contain text.
Completed practicum learning goal: “Work on a scientific data-management problem outside the scheduling thesis, learning how a laboratory team characterizes data movement and validation needs.” A tentative 12-week window in the second summer leaves a topic and host to be arranged. It does not assert a reserved laboratory place or a funded travel amount. The completed planning decision is Math/CS as the research framing, advisor review next, and a missing-letter dependency still visible.

Worked Science & Engineering plan: a physical research question
Tobin Aster’s invented doctoral direction concerns how pore structure affects fluid transport in a specified class of porous materials. HPC simulation is the means of investigating that physical relationship. His proposed output is scientific understanding of the material system, rather than a generally reusable compiler or runtime.
Completed track rationale: “I propose to investigate how pore connectivity changes simulated transport under defined boundary conditions. I would compare numerical predictions with suitable reference data and report where the model fails. The computational work serves a materials-science question; I am not proposing a broadly general scheduling technology.” This supports a Science & Engineering framing without inventing experiments, access to a machine, or successful validation.
Completed application dependency record:
| Component | Fictional preparation state | Next dependency |
|---|---|---|
| Research description | Question, method and limitations drafted | Advisor critique of scope |
| POS | Seven-course semester draft | Advisor review and course-availability check |
| References | Three identified; all letters pending | Referee agreement and actual portal receipt |
| Transcripts | Institutions listed; none requested | Official institutional delivery |
| Application | Draft, not final submitted | All required sections and review |
| Practicum | Second-summer learning goal drafted | Host, timing and program arrangements |
Completed POS outline: Science/engineering includes Porous Media Physics in Fall 2027 and Transport Phenomena in Spring 2028. Mathematics/statistics includes Numerical PDEs in Fall 2027 and Statistical Inference in Fall 2028. CS/computer engineering includes Parallel Programming in Spring 2028 and Scientific Data Systems in Spring 2029. An Introduction to HPC course is proposed for Fall 2028. This is a hypothetical planning outline under the same entry assumption as Iona’s, not an approved course list.
Completed practicum concept: Tobin proposes learning scalable analysis of observational data outside his porous-materials thesis. He would bring numerical-modeling experience while learning a different team’s data-quality and analysis methods. His tentative 12 weeks are divided into two weeks of orientation and problem definition, three establishing a baseline analysis, four developing and checking a bounded contribution, two documenting limitations, and one handing off the work. The allocation totals 12; it is an illustrative work plan, not an official required schedule.
Completed decision memo: “Use Science & Engineering framing because the central output is a physical research answer. Retain the same interdisciplinary coursework obligations as Math/CS. Keep practicum work separate from the thesis, and leave host availability, access, and expense arrangements unconfirmed. Do not label a proposed practicum a DOE employment offer.” The record remains useful even if the eventual host changes, because it identifies what must be learned and which assumptions require agreement.
Treat the practicum as a separate research commitment
The laboratory practicum page describes a 12-week experience, usually in summer but possible within the first two fellowship years. It places the work outside thesis studies and provides an additional allowance for extra expenses. Site practicum coordinators help fellows find hosts and address on-site questions.
That is more specific than “network at a national lab.” Plan for a different research environment, a bounded contribution, and the practical consequences of being on site. Clarify the topic, supervisor, dates, access requirements, housing and expense rules through the program’s actual arrangements before making commitments. The public overview does not establish an allowance amount or a guaranteed match with your preferred host.
A past fellow’s placement can help you understand research possibilities. It is not a current opening, proof that the same project accepts fellows, or evidence that a fellowship converts into employment. Keep a later lab job search separate, including its actual employer and hiring requirements.
Submit a complete current-cycle record
The application instructions identify January 7, 2027 for the current cohort and require the applicant to final-submit, not merely save sections. They also require three references and official transcripts supplied by institutions rather than applicants. Prepare the POS with an advisor; a current or recent advisor can help when a future graduate advisor is unavailable.
The current portal FAQ states that reference letters should arrive by January 7, 2027 at 5:00 p.m. CST, while its letter portal remains open during review. Do not treat that continued access as a new application deadline. Its transcript answer uses the same stated date and time. Check current instructions for a delayed fall transcript instead of assuming every missing document receives an extension.
Before final submission, reconcile the checklist, document receipts, research framing and course plan. Save the completed application and look for the submission confirmation; the instructions say to contact the program if it has not arrived within 24 hours. Neither a saved draft nor an email to a referee proves receipt of the required letter.
Write a one-page research rationale and assemble a POS draft for advisor discussion, with unresolved dependencies plainly marked. If doctoral study is not your intended commitment, keep that decision separate and browse recent roles on LandOffer.ai for an employment search.


