This document tracks daily progress, experiments, and reflections across C++ 20, Embedded Systems, Zephyr RTOS, and AI Infra phases.
| Period | Focus | Status |
|---|---|---|
| Aug 2025 | Repo creation + Env setup + Day 1 C++/Python tasks | ✅ Completed |
| Sep 2025 | Office transition & project deliverables | ⏸ Paused |
| Oct 25 2025 → Restart | Resuming with C++ 20 & Zephyr foundation | 🚀 Active |
- ✅
g++ 12.3 · CMake 4.0.3 · Python 3.10 · CUDA 12.6 - ✅ Docker image load/test (
nvidia/cuda:12.2.0-base-ubuntu22.04) - ✅ Jupyter Lab functional (NumPy demo ok)
- ✅ GoogleTest integrated in CMake workflow
- Refreshed: variables · refs · pointers · loops · functions.
- Practiced: ranges & lambda composition.
- Understood:
auto&&for range views + lvalue/rvalue binding. - Explored:
concepts+requires+std::span.
- Built NumPy arrays (reshape · slice · matmul).
- Ran DecisionTreeClassifier on Iris dataset → pred = true label ✅.
- Set up notebooks for broadcasting & stacking (next task).
- Docker save/load ✅.
- Git workflow (clean rebase · branch isolation) ✅.
- CMake + GTest runner validated through
scripts/build.sh.
- Deep dive C++ 20 (Concepts · Ranges · std::span · Design Patterns).
- Install Zephyr SDK + run
hello_worldon QEMU Cortex-M3. - Create custom Zephyr app
blinky/+ scripted build pipeline. - Document design insights in NOTES.md daily (keep git clean).
- Start Raspberry Pi SDK research for Phase 2.
- auto && in templates → universal reference for perfect forwarding.
- Input vs Random Ranges → why
std::ranges::sortrequires random access. - constexpr vs runtime computation in C++ 20 constexpr functions.
- QEMU → emulates Cortex-M3 hardware for Zephyr dev without board.
“Consistency outperforms intensity.”
Every hour spent here is a step away from corporate dependency and toward technical freedom.
☑ C++ 20 Snippets Builds
☑ GTest Running
☑ Python Sanity Model OK
☑ Docker Pipeline OK
☑ Zephyr Env Planned
☑ GitHub Repo Clean Structure
➡ Nov 7 2025: Confirm Zephyr hello world + concept labs completion before Week 2 start.