Manuel
Valle Torre

I specialize in designing and evaluating AI-driven educational environments. I engineer the full Interaction_Pipeline where humans and AI meet, ensuring systems are both technically robust and pedagogically effective.

This involves engineering Multi_Agent_Architectures (using DSPy optimizers) that orchestrate agent classifiers, experts, and tutors within real-world environments.

My core focus is on Context_Engineering—integrating runtime state (code, errors, objectives) into adaptive pipelines—and evaluating system performance using Granular_Trace_Metrics derived from authentic user research.

This research powers JELAI, a Jupyter-based platform I built to capture these rich interaction signals in real-time.

Previously at the Web Information Systems group (TU Delft), I worked on data-intensive systems. I balance my technical work with music, cycling, and engineering my own hot sauces.