Scientific ML researcher at LSU New Orleans building neural-operator and foundation-model surrogates for wave and flood modeling
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
The output axis is not scored yet: the underlying production metrics — total publication count, per-year output and venue mix — are not available in the current snapshot. It would measure how much this researcher publishes and where, and cannot be estimated from the two reviewed papers alone.
Impact—
No impact score is available: citation counts, h-index and per-paper influence are absent from this snapshot. The axis would measure how widely the work — both the neural-operator surrogates and the hydrologic forecasting line — is cited and taken up by other researchers.
Focus0/100
Scores 0/100 because the field distribution is a perfect 50/50 tie between two subfields — neural-operator surrogates for fluid dynamics and hydrologic surrogate modeling — leaving no dominant single field. In practice the spread is narrow: both catalog papers sit inside scientific machine learning, one on wave dynamics (DU-NO, FUNWAVE-TVD) and one on hydrologic forecasting (KG-Chronos-2, HEC-RAS). The zero reflects an even split, not scattered work across unrelated disciplines.
Collaboration36/100
Collaboration scores 36/100, drawn from a compact pool of 5 distinct coauthors. Both 2026 catalog papers place him in the last author position (not corresponding), consistent with a senior-contributor role on a small, stable team. Recurring partners include Elias Ioup, Enrique Hernandez Noguera, Md Meftahul Ferdaus, Nathan Cooper and Edward Holmberg. The pattern suggests deep, repeated work within one group rather than broad cross-institutional networks.
Activity—
The activity axis lacks a score because recent-year publication cadence and temporal distribution data are not in the snapshot. It would measure how active the researcher has been over time, such as papers per recent year and gaps in output, beyond the two 2026 catalog entries.
Reproducibility78/100
Reproducibility is the strongest recorded axis at 78/100, computed as a mean over 2 reviewed papers. DU-NO rates high readiness — its U-FNO comparison runs under capacity-matched, controlled parameter budgets — while the KG-Chronos-2 study rates medium. Evolve's reviews note both evaluate on real physical simulators (FUNWAVE-TVD waves; HEC-RAS flood events), though each is bounded by limited evaluation scope: single-site wave data and two events from one project.
Research summary
Mahdi Abdelguerfi, based at LSU New Orleans, works in scientific machine learning, with a cataloged focus on learned surrogates for expensive physical simulators. The Evolve catalog holds two 2026 arXiv papers, both with him in the last author position. The first, DU-NO, introduces a parameter-efficient Double U-Shaped Neural Operator for phase-resolving nearshore wave modeling; Evolve's review credits the 3.6M-parameter spectral-convolution hybrid with a 14.9% improvement over U-FNO on its own wave benchmark under capacity-matched controls, calling it a well-run study bounded by single-site data, and rates its reproducibility readiness high. The second, KG-Chronos-2, augments the frozen Chronos-2 time-series foundation model with knowledge-graph retrieval and gated correction for HEC-RAS surrogate flood forecasting, cutting event-balanced RMSE by 14.13% on two events from one project — promising, but thin evidence, with medium reproducibility readiness. Radar signals are consistent with this picture: reproducibility is the strongest recorded dimension at 78/100 (mean across 2 reviewed papers), collaboration scores 36/100 on a compact pool of 5 distinct coauthors, and focus scores 0 because his field distribution splits evenly, 50/50, between neural-operator surrogates for fluid dynamics and hydrologic surrogate modeling — a tie across two closely related branches of the same discipline rather than scattered work. Output, impact and activity metrics are not yet available in this snapshot.
Collaboration pattern
The collaboration footprint is compact: 5 distinct coauthors and a collaboration score of 36/100. Both 2026 catalog papers list Abdelguerfi as last author (not corresponding), suggesting a senior role within a stable group. Recurring coauthors — Elias Ioup, Enrique Hernandez Noguera, Md Meftahul Ferdaus, Nathan Cooper and Edward Holmberg — indicate repeated work with the same team across both wave-modeling and hydrologic-forecasting projects rather than many one-off partnerships.
Research keywords
Highlights
DU-NO: a 3.6M-parameter neural operator beats U-FNO by 14.9% under capacity-matched controls
KG-Chronos-2 cuts HEC-RAS surrogate RMSE by 14.13% with graph retrieval on a frozen foundation model
A coherent two-front surrogate agenda: nearshore waves and riverine floods
Both 2026 catalog papers replace expensive physics simulators with learned surrogates — a neural operator for phase-resolving nearshore waves and a foundation-model forecaster for HEC-RAS riverine flooding — with Abdelguerfi in the last author position on each, and his field weights split exactly 50/50 between the two subfields.
Timeline
- 2026DU-NO appears on arXiv: a 3.6M-parameter double U-shaped neural operator for phase-resolving nearshore wave modeling, benchmarked against U-FNO under capacity-matched controls.
- 2026KG-Chronos-2 appears on arXiv: knowledge-graph retrieval plus gated correction on a frozen Chronos-2 model for HEC-RAS surrogate flood forecasting, cutting event-balanced RMSE by 14.13%.
- 2026Both cataloged papers land in the same year, establishing a two-pronged surrogate-modeling line spanning nearshore waves and riverine flooding.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric scores rest on only 2 Evolve-reviewed papers (2026 arXiv preprints), so they may not reflect the full publication record or peer-reviewed output.
- Output, impact and activity dimensions are unscored: no citation counts, h-index or per-year productivity data were available in this snapshot.
- Metrics are computed from the provided Evolve radar snapshot; identity is anchored to ORCID 0009-0002-0932-2009, and no biographical details beyond the supplied fields are asserted.