Problem before model
Define the question, context, data constraints and cost of failure before selecting a technique.
Majestic Lab
Applied research to understand before scaling.
We explore AI, geospatial data and technology systems where technical, operational or ethical questions still need evidence before productization.
Research lead
Martín Heriberto Pérez Gómez
PhD researcher in Computer Science · CIC-IPN · Applied AI & GeoAI
The Lab connects doctoral research in geospatial artificial intelligence with applied questions in data, software systems and real-world operations.
Current emphasis includes geographic transferability, distribution shift, reproducible evaluation, negative transfer and the limits of model generalization across heterogeneous urban contexts.
Academic evidence, publications, institutional sources and ORCID are maintained separately from commercial positioning so that research claims remain independently verifiable.
Principles
The Lab builds evidence and judgment, not unqualified claims.
Explicit validation
Separate hypotheses, prototypes, observed results and claims that are not yet validated.
Responsible transfer
Move a research line toward product only when usefulness, limitations and operational feasibility are supported by evidence.
Public research lines
Current areas of exploration
These lines are presented as ongoing research, not as mature products or universally generalizable results.
GeoAI, spatial analysis and cross-region model transfer
Geographic variation, data quality, distribution shift and model generalization across heterogeneous urban contexts.
Responsible AI and verifiable systems
Traceability, human oversight, reproducibility, provenance and governance in AI-enabled systems used in sensitive operational settings.
Research collaboration in Applied AI, GeoAI or reproducible systems?
Majestic Lab is particularly interested in cross-city transfer learning, geospatial AI, research engineering, model evaluation and verifiable technology architecture.
Propose a collaborationEspañol