∇
Foundational Optimization
Developing root-free adaptive step sizes, nonsmooth projection methods, and variational inequalities with explicit theoretical guarantees and massive scalability.
Optimization · Machine Learning · Applied Mathematics
Pablo Barros
PhD candidate in Applied Mathematics architecting scalable algorithms for complex, high-dimensional environments. My work bridges rigorous theoretical optimization, stochastic decision-making, and mathematically grounded machine learning.
Research
Architecting the mathematics of intelligent systems.
My work bridges theoretical rigor and practical scalability, sitting at the intersection of mathematical optimization, multistage stochastic systems, and advanced machine learning.
ξ
Decisions Under Uncertainty
Pushing the boundaries of multistage stochastic programming and dynamic decision processes with dimension-free complexity bounds and trust-region stabilization.
λ
Mathematically Grounded AI
Applying rigorous mathematical frameworks to modern machine learning, from large-scale training dynamics to reliable LLM evaluation and enterprise-grade model tuning.
Education
Mathematical training and research appointments.
Academic background in applied mathematics, data science, operations research, optimization, and mathematical analysis.
Visiting PhD Student
Operations Research Center · Feb 2026 – Jun 2026Researched nonconvex optimization and deep learning under the supervision of Prof. Dimitris Bertsimas.
PhD in Applied Mathematics and Data Science
Feb 2025 – Dec 2026Relevant coursework: stochastic calculus, partial differential equations, optimization, measure theory, and functional analysis.
Bachelor in Applied Mathematics
GPA: 9.1/10 · Feb 2022 – Dec 2024Coursework in machine learning, statistical modeling/inference, time series, numerical analysis, probability, algorithms, and linear algebra.
Experience
Academic, industrial, and quantitative research background.
A mix of rigorous mathematical research, applied science, ML evaluation, quantitative finance, and olympiad-level mathematical teaching.
Applied Science PhD Intern
Microsoft · Jul 2026 – Dec 2026Working on LLM optimization and evaluation for Microsoft 365 Copilot Tuning. Studying model behavior, signal quality, benchmarking, and performance trade-offs for enterprise AI agents.
Spring into Quant Finance Participant
G-Research · Apr 2026Selected for a competitive residential programme focused on quantitative modeling, machine learning, and high-performance computing.
LLM Evaluation Analyst
Mercor · Nov 2024 – Dec 2025Performed rigorous model evaluation for AI organizations, analyzing reasoning quality, failure modes, and reliability. Contributed to scalable evaluation and review pipelines emphasizing consistency, error analysis, and quality decision workflows.
Quantitative Research Intern
Giant Steps Capital · Jan 2023 – Feb 2023Built models of implied volatility surfaces using Black–Scholes-based methods and statistical analysis. Prepared and processed options datasets with R and Python for research, signal exploration, and model validation.
Publications
Papers and manuscripts.
Collection of research papers, all published or under review at top journals and conferences in Optimization and Machine Learning, and drafts that are soon to be released.
Published
New operator designs for Halpern iterations with explicit rates under Hölder error bounds
P. Barros, V. Guigues, R. Behling, L.-R. Santos.
Mathematics of Computation (to appear)· arXiv:2601.14451→
Under Review
Parallelizing the circumcentered reflection method
P. Barros, R. Behling, V. Guigues, L.-R. Santos.
Best Poster Award, CariOPT 2025 · arXiv:2505.17258→
Parallel polyhedral projection method for the convex feasibility problem
P. Barros, V. Guigues, R. Behling.
arXiv:2506.15895→
→
→
Dimension-free complexity guarantees for dual dynamic programming
P. Barros, V. Guigues, J. Liang, R. D. C. Monteiro.
arXiv:2606.10203→
Bidirectional SDDP with dimension-free complexity for solving strongly convex stochastic dynamic programming equations
P. Barros, V. Guigues.
arXiv:2606.10161→
In Preparation
Tracking Time-Varying Equilibria in Decision-Dependent Stochastic Variational Inequalities
P. Barros, Z. Harchaoui, V. Guigues.
Expected submission · Sep 2026→
Regularized SDDP via trust-region stabilization for multistage stochastic programs
V. Guigues, V. Leclère, P. Barros, A. Shapiro.
Expected submission · Sep 2026→
Honors
Competition medals and research recognition.
International and national awards in mathematical competitions and optimization research.
International Mathematical Olympiad
🥈 Silver Medal · 2020 🥉 Bronze Medal · 2021
→
International Mathematics Competition for University Students
🏆 Grand First Prize · Top 9 · 2023
→
Ibero-American Interuniversity Mathematics Competition
🥇 Gold Medal · Top 1 · 2023
→
Brazilian Mathematics Olympiad — University Level
🥇 Gold Medal · Top 2 · 2023
→
I Carioca Workshop on Optimization and Applications
🏆 Best Poster Award · 2025
→
Classes
Mathematics classes, tutoring, and olympiad training.
I offer classes across mathematics: olympiads, undergraduate courses, exams, foundations, analysis, algebra, calculus, linear algebra, discrete mathematics, optimization, and related topics.
∀
Olympiads · Undergrad · Exams
Math Classes
Individual or small-group classes for mathematical olympiads, undergraduate mathematics, exam preparation, and advanced problem solving.
@
Email for availability
Class Inquiries
For tutoring, olympiad training, or exam preparation, contact me directly at pablock@mit.edu.
Teaching experience
Teaching Assistant
FGV EMAp · 2025Foundations of Mathematics; Algebra & Cryptography.
Teaching Assistant
FGV EMAp · 2024Optimization (PhD), Foundations of Mathematics, Mathematics 1, Analysis in Rn (MSc), and Discrete Mathematics.
Teaching Assistant
FGV EMAp · 2023Single Variable Calculus, Mathematics 1, Mathematics 2, and Linear Algebra.
Mathematics Teacher
GGE School · Feb 2022 – Jul 2025Trained top Brazilian students for national and international olympiads; guided two high-school freshmen to the IMO and several students to national OBM medals.
Contact
Links and contact.
Reach out for research, collaborations, internships, applied science, or quantitative research opportunities.
