Engineering educator and researcher building transparent computational tools.
I work at the intersection of scientific computing, AI/ML, thermal systems, digital twins, experimental design, and statistical quality. My public projects turn engineering questions into small, testable and inspectable software—so assumptions, data quality and evidence remain visible.
| Engineering question | Public work | What it demonstrates |
|---|---|---|
| Is the process stable and capable? | SPC Connect (v0.4.0) | X̄-R charts, run-signal review, capability analysis, versioned audit records and strict input validation |
| How can academic quality evidence stay traceable? | SIGC-UPA | Accreditation workflows, evidence history, surveys, reports and security checks |
| Is the thermal behavior understandable? | Thermal Digital Twin | Simulation, synthetic observations, parameter fitting, time constants and error metrics |
| Is the input data ready for SPC? | Engineering Data Quality | Non-destructive validation, missingness profiling and variability checks |
| How does a body cool or heat? | Heat and Mass Transfer Models | Lumped-capacitance equations, Biot checks and inverse calculations |
| Which factors should we study? | Engineering Experiment Design | Full-factorial designs, coded effects and physical-level decoding |
| How can a scientific ML experiment be checked on real hardware? | Scientific Computing ROCm and its 2D PINN dossier (v0.2.0) | Conservative CPU/CUDA/ROCm detection, automatic differentiation against an analytical heat solution, recorded device metadata and syntax-checked reproduction |
| How does a thermal model become hardware-ready? | FCI-UPA FPGA Lab | Fixed-point models, SystemVerilog RTL, simulation and vector verification |
I am developing a connected engineering-computation workflow:
design the experiment → validate the measurements → model the system → quantify error → document the runtime.
The emphasis is evidence over decoration: public repositories include focused APIs, unit tests and continuous integration where appropriate. Accelerator support is reported only when the local runtime exposes it; no hardware result is implied by a project name.
I am extending the same discipline to physics-informed and kinetic modelling: preserve measurement identities by construction, validate with grouped splits and held-out time horizons, quantify uncertainty through retraining, and verify numerical claims against released artefacts. The osmotic-dehydration study is still pre-submission; primary laboratory records are not redistributed, and authorship and data-use permissions are being resolved before public release.
- Scientific machine learning and AI-assisted engineering
- Thermal modeling and digital twins
- Reproducible Python for engineering education
- Design of experiments and process improvement
- Statistical process control and engineering data analysis
- Portable CPU/CUDA/ROCm workflows
- Academic accreditation systems and evidence traceability
- FPGA verification, fixed-point arithmetic and edge AI preparation
- Affiliation: Universidad Paraguayo Alemana (UPA)
- Location: Paraguay
- ORCID: 0009-0004-6275-3013
- Credential: Credly digital credential
Repositories · ORCID · Credly
Open to thoughtful collaboration around engineering computation, scientific ML, thermal systems and reproducible technical education.