Separating capability, causal effect, and rollout risk
A three-part evaluation model for controlled task performance, randomized online effects, and stability under progressive exposure.
Read the note →Writing & reading notes
Evaluation methods, computational studies, and technical notes.
Reproducible examples and focused analyses of agent evaluation, online experimentation, progressive deployment, retrieval, and stochastic optimization.
Agent evaluation
Agent Evaluation Playbook
Executable synthetic examples for task-contract evaluation, binary A/B analysis, non-inferiority guardrails, Wilson confidence bounds, and progressive rollout recommendations.
Inspect the repository ↗Research artifacts
Stochastic optimization for medical evacuation
A chance-constrained model with exogenous and decision-dependent uncertainty, Boolean reformulation, convexification, and computational analysis using realistic medical-evacuation data.
Article ↗ Code and data ↗Agentes de IA: Zero to Hero
Technical material covering ReAct, planning, tool execution, MCP, context management, RAG, memory, multi-agent systems, evaluation, observability, and guardrails.
Open course ↗