Ruan Felipe da Silva e Sousa

Ph.D. in Applied Mathematics

Stochastic optimization for power systems: the NEWAVE–DECOMP coupling used by the Brazilian system operator (ONS), and SDDP/EDDP variants for dimensionality and cut selection. Advisor: Prof. Bernardo Freitas Paulo da Costa.

M.Sc. in Systems and Computing Engineering

Thesis: Trajectory of Modern Generative Models — From Normalizing Flows to Flow Matching, through the unifying lens of Optimal Transport. Advisor: Prof. Adriano Maurício de Almeida Cortês, D.Sc.

B.Sc. in Applied Mathematics · concentration in Scientific Computing

UFRJ will always be my alma mater, and i learned a LOT there!

Consultant/Engineer

• Lead data-centric projects on an Oracle base with 1B+ records; manage a data-science team on Python systems processing ~3TB.
• Drove a multi-engine analytical benchmark (PostgreSQL, Hydra, ClickHouse, DuckDB) for the biometric-tables pipeline;
• Built a whole lot of visualizations, dynamic reports and tools for most of the work.
• Built an internal platform to present and organize the laboratory products, with access to the database with row level security, deployed safely through docker.
• Lab-wide restructuring: versioning/modularity/efficiency culture, agile squads, QA discipline. Mentor students and junior researchers.
• Construction of a new database model, as well as major redesign of the data pipeline. had a lot of fun building stuff!

Founding Engineer

• Designed and built an autonomous platform of AI agents for end-to-end SDLC automation (ticket → merged PR): specialized agents (implementer, reviewer, fixer, custodian) cooperating across iterative rounds.
• Multi-model cross-review (Claude + Gemini) to reduce single-model blind spots; guardrails via containerized validation with memory caps, least-privilege permissions, automatic escalation.
• Dynamic context construction, RAG, and type-aware AST mapping (ts-morph) over a large-scale codebase.
• Transaction pipeline for a personal-finance fintech: webhook handling (Pluggy), deduplication, multi-provider normalization, recurring-transaction detection, credit-card engine; PostgreSQL query/index optimization; Product Owner for part of the tenure.

Research Assistant (Scholarship)

• Developed a diffusion-based super-resolution model to enhance a reduced-order fluid-dynamics model (published at CILAMCE 2024).
• Refactored the reduced-order model from a notebook into a documented Python module with safe configuration loading.

Software Developer (Contractor)

• Built and deployed a portfolio-generation system based on non-linear optimization of a target metric for market trading;
• Integrated into an automated trading system. Details under NDA.

Undergraduate Research Fellow

• Research and maintenance of the BR-EMS biometric-table calculation systems;
• led a team building a Python data-cleaning system and automated data-quality reporting (Borb);
• Optimized the Oracle base a bunch. Did some more a year later ;) Advised by Prof. Milton Ramirez!

Undergraduate Research Fellow

• Stochastic-optimization methods in Julia (SDDP.jl, JuMP.jl) for the Brazilian hydrothermal dispatch problem, in partnership with ONS;
• Made reproducible Docker builds, built a lot of visualization, learned a whole bunch. Advised by the lovely Prof. Bernardo Freitas Paulo da Costa, who is fantastic!