PyTensor Beyond PyMC: Building LLM Inference in Python
An inspectable local-LLM stack built in Python: GGUF and safetensors weights, reusable PyTensor graphs, C/Numba/MLX execution, generation, and numerical validation.
Articles
Articles by Carlos Trujillo on Bayesian causal inference, marketing mix modeling, quasi-experiments, priors and budget optimization. Practical research notes with Python and PyMC.
Welcome to my collection of articles. Here you’ll find my thoughts, tutorials, and research on marketing science, causal inference and Bayesian methods.
Featured Articles
Can You Trust Your Quasi-Experiment? A Bayesian Framework for Auditing Time-Series Causal Estimates
A Bayesian framework using placebo tests and ROPE-based inference to audit whether your quasi-experimental causal estimates are trustworthy.
Decision-Making Under Contradictions: Robust Budget Allocation When Your Models Disagree
How to make robust budget allocation decisions when your measurement models (MMM, experiments, attribution) give contradictory advice.
From Experiments to Priors: Eliciting Informative Priors for Your Marketing Mix Model
How to translate quasi-experimental results into informative Bayesian priors for your MMM using CausalPy and PyMC-Marketing.
Bayesian Models and Risk Optimization
An article discussing the importance of causality in experiments. Talk given in PyData Berlin 2025.
No More Experiments Without Causality
An article discussing the importance of causality in experiments. Talk given in PyData DE Darmstadt 2025.
Baby Steps for Causal Discovery
An article discussing the importance of causality in experiments. Talk given in PyData Tallinn 2025.
All Articles
- PyTensor Beyond PyMC: Building LLM Inference in Python - July 2026
- Can You Trust Your Quasi-Experiment? A Bayesian Framework for Auditing Time-Series Causal Estimates - April 2026
- Decision-Making Under Contradictions: Robust Budget Allocation When Your Models Disagree - February 2026
- From Experiments to Priors: Eliciting Informative Priors for Your Marketing Mix Model - February 2026
- Bayesian Models and Risk Optimization - August 2025
- No More Experiments Without Causality - April 2025
- Baby Steps for Causal Discovery - February 2025