Machine learning & research
PyTorch, language-model fundamentals, retrieval, evaluation design, evidence synthesis, and reproducible experiments.
About
I like difficult systems more when their assumptions are visible.
I’m Martín, an Electronic Engineering and Mathematics student at the Universidad Nacional de Colombia in Manizales. I’m an official member of the Control and Digital Signal Processing Group (GCPDS), part of the LIA-UNAL research ecosystem, and build inspectable research and machine-learning systems, local-first software, and reliable engineering tools. My signal-processing and embedded work adds breadth across hardware and software boundaries.
Approach
Research thinking, production discipline.
My projects usually begin with a question that needs both theory and implementation: whether a tokenizer comparison is actually fair, whether a research assistant can preserve the passage behind a claim, whether personal analytics can work without surrendering data, or where an audio system’s hardware and firmware contracts should meet.
I turn those questions into explicit interfaces, tests, reproducible artifacts, and honest limits. I care about the behavior at boundaries—model output entering domain logic, archives crossing a filesystem parser, remote evidence becoming local data, or software controlling physical hardware.
Engineering range
Capabilities grounded in shipped work.
Software systems
Python, Rust, TypeScript, React, SQLite, browser automation, terminal interfaces, typed adapters, and local-first design.
Signals & embedded engineering
Digital signal processing, C/C++, STM32, USB audio, codec integration, electronics, and mixed hardware/software architecture.
Reliability
Strict typing, security boundaries, CI, branch coverage, fail-closed parsing, provenance, and limitations treated as product behavior.
Open-source collaboration
Work accepted beyond my own repositories.
Merged · March 2026
ANE2 spectral node-calibration runtime
I integrated a self-contained calibration runtime into ANE2-Calibration-SDR, including artifact provenance, PSD-frame calibration, campaign materialization, and integration tests. The upstream maintainer approved and merged the contribution.
Inspect the merged pull requestContact
Let’s work on something that has to be right.
I’m available for remote internships, part-time engineering roles, and focused contracts alongside my studies in Colombia, particularly across research engineering, ML systems, applied AI, and reliable software.