Juan Miguel Navarro Carranza
Juan Miguel Navarro Carranza
Embedded Files
AI Research Engineer with a Stanford Ph.D., building hybrid AI systems that combine foundation models with structured reasoning and rigorous evaluation. My background spans LLM robustness testing, probabilistic modeling, and high-performance computing, designing scalable pipelines for experiments, simulations, and multi-terabyte data systems that support trustworthy deployment.
Research & Interests
Research & Interests
My research interests include large language models (LLMs), agentic AI, reinforcement learning, graph machine learning, and probabilistic risk modeling. I am especially focused on questions of robustness, trustworthiness, and scalability in AI systems. Recent work includes a NeurIPS 2025 Workshop (Oral) paper on LLM robustness under paraphrase stress tests, which developed a reproducible evaluation pipeline to probe surface-form brittleness in current models.
jm.navarro.carranza [at] gmail [dot] com
Education
Education
2019-2025
Stanford University, Stanford, California.
Ph.D. in Structural Engineering and Computer Science
2019-2021
Stanford University, Stanford, California.
M.S. in Civil and Environmental Engineering, Structural Engineering, and Geomechanics.
2011-2016
Universidad de Guadalajara, Centro Universitaro de Ciencias Exactas e Ingenierías, Guadalajara, México.
B.Sc. in Civil Engineering. Summa Cum Laude.
2014-2015
Institut National des Sciences Appliqueés de Lyon, INSA de Lyon, Lyon, France.
Exchange international student. Civil Engineering and Urban Planning.
Professional Experience
Professional Experience
2026 - Present
AI Research Engineer, Daice Labs.
Prototyping hybrid composite AI architectures that combine foundation models with neurosymbolic components and bio-inspired system design for adaptive, auditable workflows.
Developing evaluation and governance-minded tooling to stress-test hybrid systems under changing context and to support traceable, explainable decision-making.
Collaborating across engineering and product teams to translate research findings into production-ready AI solutions.
2019 - 2025
Graduate Researcher, Stanford University.
Developed and published a reproducible evaluation protocol (NeurIPS 2025 Workshop) measuring surface-form robustness of LLMs via paraphrase stress tests, using Mistral-7B and Qwen2.5-7B with deterministic 4-bit inference on NVIDIA A100.
Designed and implemented large-scale HPC pipelines for nonlinear dynamic simulations of reinforced concrete buildings, scaling to hundreds of thousands of earthquake simulations on Stanford’s Sherlock cluster.
Built a modular Python data ecosystem integrating DuckDB (relational metadata) and HDF5 (time-series arrays), enabling fast querying and analysis of multi-terabyte structural simulation datasets.
Parametric calibration of the Reinforcement Ductile Fracture Model (RDFM), integrating mechanics and ML to predict reinforcement fracture under low-cycle fatigue of reinforced concrete shear walls; validated with 23 laboratory experiments.
Applied machine learning methods (Bayesian inference, probabilistic risk modeling) to quantify collapse fragility, resilience metrics, and post-earthquake degradation of building systems.
Produced visualization and analytics tools (Matplotlib, Jupyter, HPC logging) for real-time monitoring and reproducibility of simulations across thousands of jobs.
2023
Teaching Assistant - CEE 280 Advanced Structural Analysis, Stanford University.
Mentored students in object-oriented software design for structural analysis engines, guiding projects from architecture to implementation.
Led tutorials on structural analysis methods and programming skills, increasing student performance and engagement.
Provided detailed feedback on code and analysis workflows, reinforcing best practices in version control, modularity, and reproducibility.
2016 - 2019
Structural Engineer, Balderrama & Silos Ingeniería Estructural, S.C.
Co-led structural design of a $500M industrial facility and multiple commercial/residential projects, delivering structural systems from analysis through construction.
Developed custom algorithms and automation tools for analysis/design of structural elements, improving team efficiency and reducing manual calculation time.
Directed a team on the design of a 5-story residential building in Mexico City and the data center building for Megacable in Guadalajara, Mexico.
Managed structural project workflows, coordinating engineers to ensure delivery under tight deadlines.
Publications
Publications
Navarro Carranza, J.M., Navarro Carranza, J.M., LLMs Show Surface-Form Brittleness Under Paraphrase Stress Tests. NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle: Benchmarks, Emergent Abilities, and Scaling. 2025 [Oral] [Paper]
Navarro Carranza, J.M., Effects and implications of longitudinal reinforcing steel bar fracture due to low-cycle fatigue on reinforced concrete shear walls performance and post-earthquake safety. Ph.D. Dissertation, Stanford University, 2025. https://purl.stanford.edu/my842vn0510
Navarro Carranza, J.M., Deierlein, G.G., & Zhong, K. Simulation of longitudinal reinforcing steel bar fracture in reinforced concrete walls. Bulletin of Earthquake Engineering 23, 517–551 (2025). https://doi.org/10.1007/s10518-024-02078-6
Navarro Carranza, J.M., Deierlein, G.G. & Zhong, K. The influence of low-cycle fatigue on the collapse of concrete shear walls. 18th World Conference on Earthquake Engineering, Milan, Italy (2024).
Miranda, E., Navarro Carranza, J.M. et al. StEER-07 Jan. 2020 Puerto Rico mw6. 4 Earthquake: preliminary virtual reconnaissance report (PVRR) (2020).
Honors & Awards
Honors & Awards
Best Reviewer Award, NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle.
John A. Blume Fellowship, School of Engineering, Stanford University.
Gere Award, School of Engineering, Stanford University.
The Shah Family Fund Fellowship, School of Engineering, Stanford University.
Nancy Grant Chamberlain Fellowship, School of Engineering, Stanford University.
Dionisio Garza-Medina Fellowship, School of Engineering, Stanford University.
Claudio X. González Fellowship, School of Engineering, Stanford University.
Fulbright Garcia-Robles Scholarship, U.S.-Mexico Commission for Educational & Cultural Exchange, COMEXUS.
Academic Excellence Award, Colegio Metropolitano de Ingenieros Civiles de Jalisco, A.C., COMICIJ.
Distinguished Academic Performance, Ceremonia de Reconocimiento a Estudiantes Sobresalientes, Universidad de Guadalajara
México Francia Ingenieros Tecnología, MEXFITEC, Secretaría de Educación Pública, SEP.
Talks & Conference Presentations
Talks & Conference Presentations
2025 NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle - "LLMs Show Surface-Form Brittleness Under Paraphrase Stress Tests" [Oral]. San Diego, U.S.
2025 Blume Center 50th Anniversary Celebration [Poster]. Stanford, California.
2024 18th World Conference on Earthquake Engineering [Oral]. Milano, Italy.
2022 NHERI 2022 SimCenter Symposium [Poster]. Texas Advanced Computing Center (TACC). Austin, Texas.
2022 Blume Center and SURI Affiliates and Alumni Meeting [Poster]. Stanford, California.
2022 Berkeley/Stanford Computational Mechanics Festival (CompFest) [Oral]. Stanford, California.
Services
Services
Member:
2023 Fulbright Selection Committee.
Reviewer
International Conference on Learning Representations (ICLR), 2025.
International Conference on Artificial Intelligence and Statistics (AISTATS), 2025.
Conference on Neural Information Processing Systems (NeurIPS), 2024.
International Conference on Machine Learning (ICML)
NeurIPS LLM Evaluation Workshop, 2025 [Best Reviewer Award].
NeurIPS UniReps: Unifying Representations in Neural Models, 2023, 2024, 2025.
ICML Workshop on Theoretical Foundations of Foundation Models (TF2M), 2024.
ICML Workshop LXAI, 2023.
ICML Workshop on Neural Compression: From Information Theory to Applications, 2023.
ICLR Workshop on Sparsity in LLMs (SLLM), 2025.
Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL) LatinX in Natural Language Processing Research Workshop, 2024.
Journal of LatinX in AI Research (JLXAI).
LatinX in AI Supercomputing Network Cohort II (LXAISN), 2024.
Page updated
Google Sites
Report abuse