David C. Del Rey Fern'andez

Research profile

Research summary

David C. Del Rey Fernández is a researcher at the University of Waterloo whose profile is name-resolved from public bibliographic data (OpenAlex identity; ORCID 0000-0001-6946-8523). The record shows an h-index of 19 and an i10-index of 27, with output concentrated in computational mathematics: advanced numerical methods (50 works), computational fluid dynamics and aerodynamics (46), and differential equations and numerical methods (29), plus smaller clusters in turbulent flows. The current field split is even at 0.5/0.5 between AI for operations research (LLM-based optimization modeling) and nonlinear control and dynamical systems (operator learning / machine learning for control). Two 2026 arXiv papers reviewed by Evolve illustrate this pivot. 'Learning Lyapunov Operators for Nonlinear Systems' (middle authorship) takes a theory-then-practice route, proving continuity of a Lyapunov solution operator to justify Fourier Neural Operator-based stability learning, though the review flags fixable gaps in the continuity proofs and experiments covering only three 2D systems. 'SAILOR: Solver-Assisted Interactive LLM-based Optimization Recovery' (last authorship) proposes an interactive proof-of-concept system that asks users for numbers missing from LLM-translated optimization models, with evidence resting on idealized feedback and circular masking. Recent activity is solid — 13 works in the last three years and 8 recorded in 2026 — while collaboration stays compact at 5 distinct coauthors (radar collaboration score 36). Radar output and impact axes remain unscored because their underlying data is missing.

Research keywords

Highlights

• Theory-first Lyapunov operator learning via Fourier Neural Operators

  In 'Learning Lyapunov Operators for Nonlinear Systems' (2026, arXiv, middle authorship), Evolve's review credits a theory-then-practice framing: continuity of the Lyapunov solution operator is proven to justify FNO-based stability learning. The review flags fixable gaps in the continuity proofs and an experimental scope limited to three 2D systems, and rates reproducibility readiness as low.

• SAILOR: interactive elicitation to repair LLM-written optimization models

  'SAILOR: Solver-Assisted Interactive LLM-based Optimization Recovery' (2026, arXiv, last authorship) builds a proof-of-concept system that asks users for numbers missing from LLM-translated optimization models. Evolve calls the direction promising but notes the evidence rests on idealized feedback and circular masking, rating reproducibility readiness medium; the paper's keywords span interactive elicitation, clarification questions, active learning, benchmark construction, and AST program rewriting.

• A documented pivot from numerical PDE methods to learning-based control and AI for OR

  The bibliometric record frames both reviewed papers: topics are led by advanced numerical methods (50 works), computational fluid dynamics and aerodynamics (46), and differential equations and numerical methods (29), while the current field distribution splits evenly (0.5/0.5) between LLM-based optimization modeling and machine learning for control — an even-handed move into learning-based methods rather than an abandonment of the numerical core.

Timeline

1. 2008Earliest entry in the OpenAlex counts-by-year record: 1 work, marking the start of the visible bibliometric trail.
2. 2016Mid-2010s buildup: annual works rise to 6 in 2016 (5 in 2015 and 2017), aligning with the numerical-methods and CFD topic clusters.
3. 2019Peak year in the record: 10 works published, the highest annual count in the OpenAlex snapshot.
4. 2021Quietest stretch: output thins to 1-2 works per year across 2021-2023 (1, 2, 1).
5. 2026Strongest recent year: 8 works recorded, and both Evolve-reviewed papers appear on arXiv — the Lyapunov-operator study (middle author) and SAILOR (last author).

Papers in the Evolve catalog

• 2026-09-16

  Learning Lyapunov Operators for Nonlinear Systems

• 2026-09-12

  SAILOR: Solver-Assisted Interactive LLM-based Optimization Recovery

Frequent co-authors