A voice AI agent that's a first-person digital version of me. Built with Next.js, Pipecat, Deepgram STT, OpenAI LLM, Cartesia TTS, and deployed on Vercel and Pipecat Cloud

LaTeX source code and compiled assets for Lluc Santamaria's Curriculum Vitae and professional business card.

A lie is just a fact that hasn't been believed yet

A personal website hosted on GitHub Pages with a static landing page.

Design and implementation of a divergence check algorithm for recursive Match Types in the Scala 3 compiler

AI planner that auto-schedules a streaming watch-list in PDDL — handles prerequisites, simultaneous episodes, and daily time budgets, solved with Metric-FF

Rule-based Knowledge-Based System (CLIPS) that plans personalized, multi-day itineraries through the Museo del Prado using an OWL ontology of 75 artworks and heuristic classification

Local-search solver (Hill Climbing vs. Simulated Annealing) for an NP-hard package→shipping assignment problem, minimizing cost and maximizing customer happiness

A Python implementation of the algorithm for the leximin core imputation of the assignment game, based on Vazirani (2025)

Academic documentation for my Bachelor's Thesis (TFG) at UCI / FIB-UPC

Optimizes supermarket shelf layouts by maximizing adjacent-product similarity (NP-hard max-weight Hamiltonian path), using exact, greedy, and 2-approximation algorithms built from scratch. Java + S…

Four reinforcement learning algorithms — Value Iteration, Direct Estimation, Q-Learning and REINFORCE — implemented, swept and compared on the Cliff Walking environment, with the learned policies v…

Interpreter of a simplified version of J (derived from APL) implemented in Python using ANTLR and NumPy.

Discrete choice models (multinomial → nested → mixed logit) for transport mode choice in London, with policy forecasting, value of time and elasticity analysis. Built in Python + Biogeme. EPFL cour…

Binary classifier predicting coronary heart disease risk in 328k patients, implementing logistic regression and gradient descent from scratch in NumPy with cross-validation, missing-data handling, …

Built for the Y Combinator 2026 hackathon, ClearPath is a voice agent that fills medical intake forms with full browser observability

Explored Reinforcement Learning algorithms in a partially observable, multi-agent environment. Focused on Joint-Action Learning leveraging Game-Theoretic solution concepts to study agent coordinati…