Alexander Krol - Portfolio

Published.

3 papers in preparation, 1 acknowledged.

Awards.

U.S. National Chemistry Olympiad Finalist – Coca-Cola Scholars Semifinalist – Harvard Book Award – Ralph E. Oesper Award for Chemistry – Governor's Award for Excellence in Environmental Protection Research – Governor's Edison Award for Advanced or Alternative Energy – Highest Achievement in Organic Chemistry (UC) – ACT 36 / 36.

Bilingual.

Native and fluent in both English and Spanish.

What have I worked on?

Wrist X-rays measured in seconds.

A transformer pipeline that automates radial inclination, volar tilt, and ulnar variance from PA and lateral wrist films – 83% faster than manual radiology.

Rational agents that learn to bet and search.

Reinforcement learning for sequential decision-making in exhaustible environments. How risk-averse strategies emerge from constraint, not reward.

Zero-shot instrument separation.

Prompt-conditioned stem separation in the AudioLDM2 VAE latent space – latent mel-spectrogram diffusion paired with STFT masking. Text or audio queries isolate arbitrary instruments without per-source training.

Programmable nanoparticle synthesis.

High-precision programmable syringe pump for automated nanoparticle synthesis. 9× throughput, plus a classifier across morphology and size.

Drug-target binding that explains itself.

Cross-modal transformer over ChemBERTa and ESM-2. CI 0.882 on DAVIS; the cross-attention map autonomously localizes DFG, P-loop, and ATP binding sites without binding-site supervision.

Intestinal stem cells under PPARα stress.

PPARα overexpression and its effect on intestinal stem cell differentiation. qPCR, IHC, cell culture, immunofluorescence at Cincinnati Children's.

Targeting Ewing's sarcoma.

Protein expression in Ewing's sarcoma cell samples; identifying treatment targets and reviewing nIGF-1R's role in tumorigenesis.

The big picture.

Across research positions, the same throughline: take a measurement that is slow, expensive, or noisy, and replace the bottleneck with a model that is fast, repeatable, and interpretable.
The models I ship are evaluated on the metric that actually matters – radiologist hours, synthesis throughput, cold-target generalization, and binding-site recovery without supervision.
Currently at Duke (B.S. ECE with ML concentration – B.S. Chemistry), Class of 2027. Open to research, internships, and collaborations.

Coursework.

Machine Learning & Stats

Applied Probability for Statistical Learning Applied Machine Learning Neural-Net-Based Large Language Models Computational Approaches to Human Language Probability Theory Statistics

Mathematics

Linear Algebra Multivariable Calculus Calculus I & II Ordinary & Partial Differential Equations Stochastic Calculus

Electrical & Computer Engineering

Data Structures & Algorithms Fundamentals of Electrical & Computer Engineering Microelectronic Devices & Circuits Signals & Systems Engineering Design & Communication

Chemistry & Physics

Physical Chemistry I: Quantum Mechanics Physical Chemistry II: Statistical Mechanics & Thermodynamics Analytical Chemistry Organic Chemistry I & II General Chemistry Mechanics Electricity & Magnetism

Liberal Arts

Academic Writing Introduction to Philosophy Sociological Inquiry Medical Sociology Microeconomics Macroeconomics Introductory Psychology Advanced & Intermediate Spanish American History I & II Genetics and Molecular Biology

Selected work – published, in review, and in preparation.

2026

Normative bet-spreading strategy in an exhaustible environment.Krol, Kim et al. – investigates rational risk-taking in resource-limited environments via reinforcement learning.

In preparation – First author

2026

Correlated search strategies and collective efficiency in exhaustible environments.Kim et al. – examines how correlation among parallel agents affects collective search performance.

In preparation – Co-author

2026

Scalable synthesis and SERS-based detection of nanomaterials.Strobbia et al. – automated pipeline for synthesis and chemical detection of engineered nanoparticles.

In preparation – Co-author

2023

IGF-1R targeting in cancer – does sub-cellular localization matter?Soni et al., Journal of Experimental and Clinical Cancer Research, 2023. Acknowledged.

JECCR – Acknowledged

Frequently asked questions.

What is your background?

I am a Spanish-American dual citizen and a Duke master’s-track student in Machine Learning, with a strong passion for applying ML across a wide variety of fields. I am self-motivated and love picking up a new field, learning it, and contributing to it.
What programming languages do you have experience in?

PythonRMATLABJavaJavaScriptC / C++SQLHTMLCSSLaTeX
What machine learning libraries have you used?

PyTorchTensorFlowJAXHuggingFacescikit-learnNumPyPandasSciPyMatplotlibOpenCVReact
What machine-learning techniques have you worked with?

Deep LearningReinforcement LearningTransformersAttentionComputer VisionNLPDiffusion ModelsClassical MLGANsGradient BoostingLSTMCNNRNNViTBERTGPTFine-TuningPPORAGLLMs
What are your musical interests?

Classical piano at the Cincinnati Conservatory of Music – repertoire spanning Liszt, Chopin, Beethoven, Rachmaninoff, and Scriabin. Spanish classical guitar – Tarrega, Llobet, Rodrigo. Vocal ensemble. Steel drum band.
What other interests do you have?

I love rock climbing and recently hit my first 7c / 5.12d on lead! I also love playing and watching tennis – especially Alcaráz – running, and teaching: I have started two separate ESL programs with over 300 participants total. I am also a huge fan of chess – I started a competitive league in my hometown.

Alexander Krol Rojas

I build machine-learning systems across medicine, agentic AI, materials, and audio.

What have I worked on?

The big picture.

Coursework.

Selected work – published, in review, and in preparation.

Frequently asked questions.

What is your background?

What programming languages do you have experience in?

What machine learning libraries have you used?

What machine-learning techniques have you worked with?

What are your musical interests?

What other interests do you have?

Let's build the next thing.