# Employee Spotlight: Enrique de Leon Hicks
# Source: https://quadsci.ai/blog/enrique-de-leon-hicks-spotlight
# Format: RAG-optimized full article text with YAML frontmatter

---
content_type: blog_post
title: "Employee Spotlight: Enrique de Leon Hicks"
url: https://quadsci.ai/blog/enrique-de-leon-hicks-spotlight
date_published: 2026-08-25
category: Culture
author: QuadSci Team
---

Enrique joined QuadSci this past spring, carrying an ML engineering toolkit built on large-scale document understanding, a Fulbright-supported Harvard master's, and a research foundation that began at Tecnológico de Monterrey. He currently manages the critical feature engineering layer for QuadSci's models, focusing on the rigorous task of ensuring every signal is verifiable and concrete. Here, he discusses why explainability felt like the right choice, his definition of impactful intelligence, and his openness to any padel invitation.

## Why QuadSci?

When I first looked at the problem QuadSci is solving, it didn't feel like a bet, it felt obvious. Every company selling software already knows customers churn or grow, but almost none of them can say why in language a person can act on. Pairing prediction with explainability, tracing a signal back to something real in the customer's data instead of a chart full of coefficients, is the kind of thing that looks obvious only once someone's already done it. That's the pull for me: the right idea, arriving at the right time.

## What's your background?

I studied informatics at Tecnológico de Monterrey, and published my first paper as an undergrad on an attention-based method for algorithm selection. From there, I spent the next several years as an ML engineer, fine-tuning and deploying transformer-based models for large-scale document understanding: building datasets and taxonomies, running the evaluation work that decided whether a change actually shipped, and going deep enough into fine-tuning and long-context techniques to know where they break. I went back to Harvard for a master's in data science on a Fulbright, TA'ing a graduate deep learning course along the way. My focus split between scalable, reliable distributed systems and deep learning, from the traditional statistical and ML foundations through the latest work on LLMs and agents, along with the statistics of how you actually validate whether a model holds up. That gave me the skills I think this work actually needs. I joined QuadSci shortly after finishing this past spring.

## What are you building right now?

Right now I own the feature engineering layer that every model at QuadSci sits on top of: turning raw customer usage data into the inputs models train on, at a scale and consistency where one small error quietly biases every prediction built on top of it. The work is less about any single model and more about building engineering discipline around features, making them deterministic, reproducible, and traceable back to the exact source value that produced them. That last part matters most to me. QuadSci's whole differentiator is turning a feature into a signal a person can trust, and that promise only holds if the layer underneath it is rigorous enough to make the trace-back exact, not approximate.

## What does accurate and useful intelligence look like to you?

I used to run a benchmark whose entire job was to check whether a metric that looked better on paper actually made results better for a real person. That gap, between a number moving and a person's outcome changing, is the only thing I actually trust. A model can be perfectly calibrated and still be useless if nobody downstream can act on what it's telling them.

## One thing people should know about you

I've played piano and tennis since I was a kid, and tennis is still my favorite thing to do outside of work. I'll also take a game of squash or padel whenever someone's up for it.