Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

GPU Optimization for Python (NVIDIA)

Moves CPU-bound scientific Python to NVIDIA GPUs and proves the port is both correct and genuinely faster.

Dev & CodingAdvanced★ 33,030⑂ 3,248AI score 8/10Last updated: Aug 9, 2026

What it does

  • Treats GPU acceleration as an evidence-driven optimization: baseline first, profile, then decide whether porting is worth it.
  • Maps your workload to the smallest suitable layer — NumPy→CuPy, pandas→cudf.pandas/cuDF, scikit-learn→cuml.accel/cuML, NetworkX→nx-cugraph, scikit-image→cuCIM, Faiss/k-NN→cuVS, file I/O→KvikIO, simulation→Warp/Newton, low-level→RAFT.
  • Enforces correct GPU benchmarking (CUDA events, cupyx.profiler.benchmark, Nsight Systems/Compute) instead of timing async enqueues.
  • Validates numerics against CPU with explicit tolerances, reports recall@k for approximate search, and flags silent CPU fallbacks.
  • Flags sunset/archived projects (cuxfilter, cuSpatial) and suggests modern replacements.

Who it's for

  • Researchers, data scientists, and ML engineers working with large arrays, dataframes, or graphs.
  • Teams evaluating RAPIDS for the first time and unsure which library fits.
  • Anyone with existing GPU code who wants trustworthy speed and accuracy measurements.

Examples

  1. "My 50M-row pandas groupby is too slow" → start with cudf.pandas accelerator mode, fall back to native cuDF, and benchmark end-to-end including transfers.
  2. "Speed up my RAG retrieval built on sklearn NearestNeighbors" → port to cuVS and report recall@k versus exact search.
  3. "NetworkX PageRank takes hours" → enable the nx-cugraph backend with near-zero code change and inspect logs for CPU fallback.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/K-Dense-AI/scientific-agent-skills/HEAD/skills/optimize-for-gpu/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/optimize-for-gpu folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/optimize-for-gpu/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/K-Dense-AI/scientific-agent-skills.git /tmp/sas && mkdir -p ~/.claude/skills && cp -r /tmp/sas/skills/optimize-for-gpu ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Confirm you have an NVIDIA CUDA-capable GPU on Linux or WSL2 with Python 3.11+.
  2. Clone the repository: git clone https://github.com/K-Dense-AI/scientific-agent-skills.git
  3. Copy the skill folder: mkdir -p ~/.claude/skills && cp -r scientific-agent-skills/skills/optimize-for-gpu ~/.claude/skills/
  4. Verify the references/ subfolder came along — it holds the per-library API guides the skill reads before writing code.
  5. Restart Claude Code and ask something like "optimize this NumPy pipeline for my GPU" to trigger the skill.
  6. Install the GPU packages (CuPy, RAPIDS, Warp, etc.) matching your CUDA version, following references/installation.md.
View source on GitHub ↗License: MIT