Welcome — this repository is the special profile README for my GitHub and serves as a lightweight portfolio landing page. Below you'll find a featured project, highlights of my work, tech stack, and contact links.

A modern take on the classic Battleship game with machine learning components. Battleship-ML demonstrates applied ML in a game environment, a configurable CLI, and an experimental web demo via WebAssembly.

• Highlights: AI opponent (reinforcement learning / heuristic hybrids), modular game engine, training pipeline, and an optional browser demo (WASM).
• Tech: C++, Python (training), WebAssembly, PyTorch/TensorFlow (or your preferred ML lib), build tooling (CMake/emscripten).

Screenshot / Demo

Quick start (example)

# Clone this repo (if Battleship code is in a subfolder, adjust accordingly)
git clone https://github.com/JohnScarrow/JohnScarrow.git
cd JohnScarrow

# If the Battleship project is in `battleship-ml/`:
cd battleship-ml
# Build/run (example; replace with your project's build commands)
mkdir -p build && cd build
cmake .. && make
./battleship

# Or run the training script (example)
python3 train.py --config configs/train.yaml

Play the demo

Note: GitHub sanitizes README.md HTML and disallows embedding active content like iframe. Because of this you cannot directly play the demo inside the README itself. Use the links below to open the demo or visit a Pages-hosted iframe page.

• Direct demo URL: https://johnscarrow.github.io/battleship-ml/
• Open GitHub Pages iframe page: https://<your-username>.github.io/<repo>/ after enabling Pages (I added a docs/index.html that embeds the demo; see below).

You can replace <your-username> and <repo> with your GitHub username and this repository name if you want the iframe page to live under this repo (for example https://johnscarrow.github.io/JohnScarrow/).

If you want to run the Battleship-ML tuner locally with GPU acceleration (to play many more games concurrently), follow these steps.

• Prerequisites:

  • NVIDIA GPU with a compatible driver installed.
  • CUDA toolkit (nvcc) installed and on your PATH (CUDA 11+ recommended).
  • Standard build tools (g++, make, etc.).

• Quick checks:

  • nvidia-smi — confirm the GPU is visible and drivers are loaded.
  • nvcc --version — confirm CUDA toolchain is installed.

• Build the CUDA-enabled tuner (from the battleship-ml/ folder):

cd battleship-ml
./scripts/build_tuner_cuda.sh

This creates a tuner binary that will use CUDA if available.

• Running for high concurrency:
  • Single-process (use threads to increase concurrency inside one process):

./tuner games=10000 threads=32 mcIterations=1600 > results.csv

- Multiple processes (useful for many independent runs or multi-GPU):

# 4 processes sharing the same GPU (they will contend for GPU resources)
for i in 1 2 3 4; do
	./tuner games=2000 threads=8 mcIterations=800 > results-$i.csv 2>&1 &
done

# Pin one process to GPU 0 and one to GPU 1 on multi-GPU systems
CUDA_VISIBLE_DEVICES=0 ./tuner games=5000 threads=8 mcIterations=1200 > out-gpu0.csv &
CUDA_VISIBLE_DEVICES=1 ./tuner games=5000 threads=8 mcIterations=1200 > out-gpu1.csv &

• Monitoring & tuning:

  • Monitor nvidia-smi (e.g. watch -n1 nvidia-smi) to observe GPU utilization and memory.
  • Increase mcIterations to give the GPU more work per Monte-Carlo decision (reduces host/GPU overhead).
  • Adjust threads to match CPU cores and how the workload pipelines to the GPU.

• Troubleshooting:

  • If the binary falls back to CPU, confirm you built with the provided CUDA script and that nvidia-smi and nvcc succeed.
  • For driver/CUDA mismatch, update your driver or use a CUDA toolkit version compatible with your driver.

• Responsible Consumption — React + Express. A journaling app for mindful consumption. Source available in separate repo(s).
• PlaylistFor.ME — JavaScript. Playlist generator using external music APIs.

• 🎓 First-year Computer Science student at North Idaho College
• 💻 Full‑stack development, AI integration, and systems engineering
• 🛠 Projects: game engines, ML experiments, autonomous systems, and cloud tooling

• Languages: C++, Python, JavaScript, Arduino/C
• Frameworks / Tools: React, Express, PyTorch / TensorFlow, Electron, Emscripten
• Cloud / DevOps: AWS, GitHub, Docker (learning)

• GitHub: https://github.com/JohnScarrow
• LinkedIn: (https://www.linkedin.com/in/johnascarrow/)
• Portfolio: https://johnscarrow.github.io/

If you'd like, I can:

• add live demo hosting via GitHub Pages or gh-pages branch
• create a minimal docs/ site for the WASM demo
• add screenshots and a short walkthrough for Battleship-ML

⭐ Thanks for stopping by — want me to set up GitHub Pages next? (I can scaffold a docs/ site and add a demo.)