Worked on enhancing the Nixtla/neuralforecast repository by improving its dependency management to support scientific computations. Focused on Python package development, the work involved adding SciPy as a runtime dependency and updating the project configuration to ensure its availability during runtime. This technical approach addressed the need for accurate numerical routines and reduced the risk of runtime failures caused by missing packages. By enforcing consistent environments for both development and production, the changes strengthened reproducibility and stability for numerical experiments and benchmarks. The contribution centered on Python and dependency management, directly supporting the reliability of scientific workflows within the project.