Independent research · Applied systems

Romel Lira

National security, federal growth, analytical systems, and applied AI.

I work on systems for consequential decisions—focused on how evidence is handled, how reasoning can be traced, and where probabilistic models stop being defensible analysis.

Selected work Professional perspective

Professional focus

Practice first. Systems second.

The work is shaped by high-stakes analytical environments, institutional constraints, and the need to make complex judgments explainable to other people.

  • High-stakes analysis National-security analysis and decision support inform the questions I choose to examine.
  • Federal growth A federal-growth perspective keeps technical ideas connected to institutions, adoption, and practical constraints.
  • Applied AI experimentation Working software tests ideas about provenance, evidence, reproducibility, and failure modes.

Selected work

Systems that make reasoning inspectable.

NTIS is the flagship project within a broader interest in analytical systems, decision support, and responsible uses of machine perception.

Flagship research project · Experimental · Pre-commercial

Non-Trivial Intelligence Systems

NTIS is a research framework for examining how analytical reasoning changes across time and pressure—and for preserving the path from evidence to judgment. It is not currently offered as a commercial product or service.

Probabilistic perception; deterministic analysis.

Read about NTIS Technical documentation
The problem

Plausible learned output can look authoritative without showing a defensible chain from evidence to conclusion.

The approach

Trace rhetorical structure, stylometric change, evidence signals, and analytical variation over time.

The standard

Make provenance, reproducibility, uncertainty, and failure modes visible while keeping human judgment explicit.

Professional perspective

AI is an analytical instrument, not an oracle.

Useful decision support has to survive scrutiny. That means separating perception from analysis, documenting evidence boundaries, and making the transformation from source material to judgment reproducible.

My work explores that boundary through systems design and practical experimentation: what probabilistic models can surface, what deterministic methods must verify, and where people need to remain visibly responsible for the decision.

Miso, the Nontrivial Ventures dragon-cat mascot

A deliberate signature

Miso keeps the rigor from becoming generic.

The dragon-cat is a reminder that technical seriousness does not require sameness. Signal, not soup—with a little personality left intact.