fortriage
AI-powered prioritized dashboard for legacy Fortran systems. Combines static analysis, risk prioritization, interactive visualization and a conversational assistant to decide what to modernize first.
Adolfo J. Cardozo S. · AC Scientific Computing
Everything I have public on GitHub, in one place: tools to understand and modernize legacy Fortran code, infrastructure for integrating language models, Python utilities and practice projects. Code you can review before working with me.
Featured projects
Filter by area. Each card links to the repository, and to the demo or package when there is one.
AI-powered prioritized dashboard for legacy Fortran systems. Combines static analysis, risk prioritization, interactive visualization and a conversational assistant to decide what to modernize first.
Static analysis toolkit for real-world, mixed Fortran code (F77/F90/F95), right where general-purpose parsers fail. Resumable 19-step pipeline with detailed logging.
Provider-agnostic LLM access layer. Ports-and-adapters design: LiteLLM, Any-LLM or native SDKs as interchangeable adapters, plus a policy, context and orchestration core.
Read and write data with Fortran FORMAT syntax from Python: fixed and free formats, type inference and formatted output.
Find and manage every Python virtual environment on your disk: venv, conda, poetry, uv, pipenv, hatch and more. Fast scanning, parallel too. Windows, Linux and macOS.
AI engineering projects (LLM APIs, RAG, agents) taken beyond the course material: object-oriented refactoring and open-model integration.
Classic Snake as a standalone browser game. Lightweight, no dependencies, deployed on its own subdomain.
No featured projects in this category; see the list below.
Practice work, experiments and earlier projects. Also on GitHub.
Why choose me
I have developed mathematical models, implemented them in high-performance Fortran and run them on HPC clusters. That is why I understand both the code and the problem it solves.
I can tell when a result “runs” but makes no physical sense.
I explain limits and complexity up front, and only commit to what I can deliver well.
This portfolio is the proof: how I structure, test and document, visible before you hire me.
One to three projects at a time. You talk directly to the person writing the code, in English or Spanish.
How I work
Testimonials
“We had a 1990s Fortran model nobody dared to touch. Now it runs as a Python library, with tests and the same numerical results.”
“What I valued most was the clarity: from the proposal on, we knew what would be delivered, what wouldn’t, and why.”
“He added an LLM assistant to our analysis without locking us into one provider. We moved to a local model without rewriting anything.”
FAQ
Contact
Tell me in a few lines what problem you want to solve. I will reply with specific questions or a proposal.