Research

Applied research, engineering that operates

Menlo Tech develops and applies computational methods to complex operations: mathematics, AI and agents, with labeled evidence and human oversight.

Applied AI

Our starting point is the business problem. Language and learning models are part of the solution; the decision comes from explicit mathematics, and execution from agents that respect policies defined by the company.

Eight research areas

Agentic AI

Agents that plan, use tools and execute within limits: coordination, memory, state and human approval.

Published indicator

See the Subway Brasil case

Mathematical Optimization

Linear and integer programming, constraints and resource allocation applied to publishing, collections and routing.

Illustrative example

See models in the RE/MAX case

Decision Intelligence

Decision theory, Bayesian inference and simulation to make choices explainable and testable.

Illustrative example

See models in the Clave case

Knowledge Systems

Knowledge graphs, entity resolution and provenance: every fact with origin and relations.

Illustrative example

See the knowledge graph

Predictive Intelligence

Demand and payment forecasting, probability calibration and Markov chains.

Proposed architecture

See forecasting in the RE/MAX case

AI Evaluation

Calibration, drift, testing and monitoring: claim only what is measured.

Proposed architecture

See evaluation systems

Distributed Systems

Event-driven execution, state machines and parallel orchestration.

Proposed architecture

See the infrastructure

Human-AI Collaboration

When AI acts alone, when it asks for approval and how a person takes over the decision.

Published indicator

See governance in Subway

How the five disciplines connect

Research philosophy

Mathematical rigor

Every decision has a model behind it, with declared variables, objective, constraints and limits.

Labeled evidence

Each claim states whether it is published, illustrative, simulated, proposed or not yet published.

Human oversight

AI executes within policies; when a rule is exceeded, a person decides.

Engineering that operates

Research matters when it becomes a system that runs, is observable and can be audited.

Research papers and technical publications

We have not yet published research papers with experimental results. This section does not present experiments that were not performed.

What exists today are technical publications based on the documented projects:

Algorithm tour: RE/MAXInteractive tour · Open
Algorithm tour: ClaveInteractive tour · Open
Algorithm tour: Subway BrasilInteractive tour · Open
RE/MAX case: mathematical formulation and agentsCapability technical article · Read
Clave case: financial mathematics and confidenceCapability technical article · Read
Subway Brasil case: priority and operationCapability technical article · Read

Publication template

Every future publication will follow this structure. It is a template: it contains no experiments or results.

AbstractThe problem, the method and what was observed, in a few lines.
Problem statementThe decision or operation to improve and why.
Research hypothesisWhat is expected, stated so that it can be refuted.
Mathematical formulationVariables, objective, constraints and assumptions.
ArchitectureHow the method becomes a system.
MethodologyHow the hypothesis will be tested.
Experimental setupData, parameters, baselines and measures.
ResultsWhat was measured, with uncertainty and comparison.
LimitationsWhat the study does not show.
Enterprise applicationsWhere the result can be used and under what conditions.

Editorial structure template. No experiment is claimed here.

Discuss a research partnership

If you have an operational problem with real data and constraints, tell us.

Talk about research