ScDomi/ScDomi：scryx 的 AI 实验与数据系统公开项目页
ScDomi/ScDomi
GitHub 上的 ScDomi/ScDomi 是 scryx 的公开项目页，汇集 AI 实验、数据系统与研究片段。公开作品包括德国可再生能源数据分析与太阳能预测、PyTorch 与 Gymnasium 实现的 DQN CarRacing 智能体、EMNIST 深度学习展示，以及数学与计算机基础项目。其研究涉及用于大学数学辅导的 prompt-moderated LLM。
scryx · rules are defaults, not truth
AI experiments · data systems · research fragments · externalized brain clutter
surface of the machine.
rules are defaults, not truth.
This profile is not meant to be a complete archive. It is a curated surface of the things my brain keeps circling: models, learning, data, systems, interfaces, rules, weird patterns, and the small artifacts that make them inspectable.
The motto is simple: rules are defaults, not truth. Some rules create clarity. Some protect people. A lot of them are just inherited settings with better branding.
I like opening the surface and looking for the machinery underneath.
brain map
┌──────────────┐
│ learning │
│ systems │
└──────┬───────┘
│
┌─────────────┼─────────────┐
│ │ │
models data traces interfaces
│ │ │
└──────► artifacts ◄───────┘
│
proof over posture
- AI / ML — experiments, evaluation, computer vision, reinforcement learning
- Data — visual explanations, forecasting, dashboards, pattern-finding
- Research — AI for learning, LLM tutoring, Learning Analytics, mathematical reasoning
- Systems — local workflows, automation, context-heavy tools, useful weirdness
- Interfaces — making technical things feel less dead
main surface
scryx
scryx is the public shell around the same idea: a readable cut through my head, not a portfolio template wearing a dark hoodie.
AI, data, research, aesthetics, systems, unfinished thoughts that became tools — arranged as a site instead of rotting in folders. The visual direction is deliberately neural: spine, brain, particles, project tiles, and the surface of the machine. motto.
what scryx is trying to do
- represent the brain/system metaphor behind the work
- keep the motto visible: rules are defaults, not truth
- make AI/data/research projects discoverable without flattening them into a CV
- connect the technical layer with the visual/personality layer
- stay strange, but still navigable
public artifacts
Renewable Energy AnalyticsGermany energy data, renewable growth and solar forecasting. Python · pandas · forecasting · visual explanation |
DQN CarRacing AgentPixels in, steering out. A small agent learning where not to die. PyTorch · Gymnasium · DQN · reward traces |
EMNIST Deep Learning ShowcaseHandwritten characters, CNN experiments and evaluation that does not hide behind vibes. PyTorch · transfer learning · Optuna · confusion matrices |
Math & CS FoundationsThe non-flashy layer: algorithms, logic, small simulations, constraint solving. Java · Prolog · C++ · Python · foundations |
research signal
Published research around prompt-moderated LLMs for tertiary math tutoring — AI for learning, but with evaluation and actual educational context instead of pure prompt-demo energy.
- ACL Anthology profile
- scryx for the broader public-facing layer
stack I reach for
entry points
- scryx — public site, project shell and identity layer
- Renewable Energy Analytics — data analytics / forecasting case study
- DQN CarRacing Agent — reinforcement-learning showcase
- EMNIST Deep Learning Showcase — computer-vision / PyTorch showcase
- Math & CS Foundations — foundations archive
- ACL Anthology — research profile
currently: externalizing the brain — fewer things, more signal per thing
来源：GitHub · github.com