# Fernando Campos Silva Dal' Maria

> Computer Science developer and researcher focused on computer graphics, artificial intelligence, and high-performance software.

Areas: Computer graphics, Artificial intelligence, High-performance computing. Belo Horizonte · Brazil.

I build tools that turn complex ideas into clear interfaces—from deep-learning frameworks in C++ and CUDA to interactive graph and automata visualizations. My previous medical education also informs an interdisciplinary approach centered on real problems and the people affected by them.

Skills: C++, CUDA, PYTHON, PYTORCH, TYPESCRIPT, REACT, RUST, DEEP LEARNING, PARALLEL COMPUTING

## Research papers

- [Persuasive Technology for Managing the Routine of Families with Children and Adolescents with ADHD: A User-Centered Approach](https://doi.org/10.1145/3702038.3702101): IHC 2024 · ACM, DOI 10.1145/3702038.3702101
- [Support for Families of Children and Adolescents with Attention-Deficit/Hyperactivity Disorder: A Technological Solution Based on Persuasive Strategies](https://doi.org/10.5753/jbcs.2026.5670): Journal of the Brazilian Computer Society · 32:1, DOI 10.5753/jbcs.2026.5670

## Projects

- [Pineapple](https://github.com/FernandoCsm-Knight/Pineapple): A deep-learning framework written from scratch in C++, with automatic differentiation, tensors, modular layers, and parallel CPU/GPU execution with CUDA. (C++, CUDA, APRENDIZADO PROFUNDO)
- [Graphic Paint](https://graphic-paint-roan.vercel.app/paint): A TypeScript visual editor for graphs and automata, with input simulation and interactive algorithms directly on the canvas. (TYPESCRIPT, REACT, ALGORITMOS)
- [Stellar Classification](https://github.com/FernandoCsm-Knight/Stellar-Classification): Machine-learning models applied to astronomical data to classify stars, galaxies, and quasars. (PYTHON, SCIKIT-LEARN, MACHINE LEARNING)
- [ADHD Research](https://dl.acm.org/doi/10.1145/3702038.3702101): User-centered research on persuasive technology supporting the routines of families of children and adolescents with ADHD. (PESQUISA, IHC, ACM)
- [Relationship Hypergraph System (RHPG)](https://github.com/FernandoCsm-Knight/Relationship-RHPG): Organizational performance analysis system based on graphs and hypergraphs. It combines individual indicators, collaboration relationships, team participation and structural network metrics to provide contextual employee assessments. The system includes an interactive dashboard, group affinity and impact analysis, centrality metrics, performance classification and AI-assisted employee and team-fit evaluations. (NETWORKX, HYPERNETX, OPENAI, PLOTLY, MACHINE LEARNING, GRAPH THEORY)
- [Relationship Graph Interpreter](https://relationship-graph-interpreter.vercel.app/): Relationship graph analysis and interpretation system designed to identify patterns associated with nepotism. The tool examines connections between individuals and relationship structures to detect clusters, proximity patterns and potentially suspicious relationships, presenting the results through visual and interpretable analyses. (GRAFOS, DETECÇÃO DE PADRÕES, ANÁLISE DE REDES)
- [Fannar](https://github.com/FernandoCsm-Knight/Fannar): Mechanistic interpretability method based on the geometric analysis of neural network internal representations. Fannar uses kernels and reproducing kernel Hilbert spaces to model relationships between activations, parameters and gradients, enabling the analysis of the geometry formed across network layers. From this structure, the method investigates similarities, dominant modes, energy concentration and the progressive formation of the model's decision. (INTERPRETABILIDADE MECANICISTA, ANÁLISE GEOMÉTRICA, RKHS, KERNEL METHODS, DECOMPOSIÇÃO ESPECTRAL, REPRESENTATION ANALYSIS)
- [Rescue Drones](https://fernandocsm-knight.github.io/Drone-Rescue/): Interactive drone search-and-rescue simulator. It uses noisy cellphone signal measurements to estimate a missing person’s location and optimizes drone positions using Fisher information to reduce uncertainty. (DRONES, OTIMIZATION, SIMULATION)

## Résumé

- [Currículo em português](https://fernandocsdm.com/docs/fernandocsdm_cv_pt.pdf)
- [Résumé in English](https://fernandocsdm.com/docs/fernandocsdm_cv_en.pdf)
- [Currículum en español](https://fernandocsdm.com/docs/fernandocsdm_cv_es.pdf)

## Contact

- E-mail: fernandocsdm@gmail.com
- LinkedIn: https://www.linkedin.com/in/fernandocsdm
- GitHub: https://github.com/FernandoCsm-Knight
- [Send a project brief](https://fernandocsdm.com/contact)

## Pages

- [Home](https://fernandocsdm.com/): overview, areas of expertise and visitor reviews
- [About](https://fernandocsdm.com/about): background, timeline, papers and certifications
- [Projects](https://fernandocsdm.com/projects): interactive map of published projects
- [Pricing](https://fernandocsdm.com/pricing): hourly rates by type of work, adjusted per region
- [Contact](https://fernandocsdm.com/contact): encrypted contact form for project requests
- [Fannar](https://fannar.fernandocsdm.com/): model interpretability services on a subdomain: audits of trained models (trees, neural networks, vision, text, time series) with Fannar, SHAP, and LIME, fixed-scope packages, and worked examples. Full content: https://fannar.fernandocsdm.com/llms.txt

## Optional

- [Full content (EN + PT)](https://fernandocsdm.com/llms-full.txt)