chrisvoncsefalvay

Training manager for local GPU training - validate CUDA, manage GPU selection, monitor progress, handle checkpoints

Training manager for RunPod GPU instances - configure pods, launch training, monitor progress, retrieve checkpoints

Validate datasets for Unsloth fine-tuning. Use when the user wants to check a dataset, analyze tokens, calculate Chinchilla optimality, or prepare data for training.

Training manager for Hugging Face Jobs - launch fine-tuning on HF cloud GPUs with optional WandB monitoring

Generate Unsloth training notebooks and scripts. Use when the user wants to create a training notebook, configure fine-tuning parameters, or set up SFT/DPO/GRPO training.

Generate comprehensive model cards and upload fine-tuned models to Hugging Face Hub with professional documentation

Build React applications with CopilotKit and LangGraph for AI-powered collaborative editing. Creates production-ready editor applications (text, document, or node-based) with real-time AI assistance, shared state management, and agentic workflows. Use this skill when users want to build intelligent editing interfaces with AI collaboration features.

Scaffold a native-looking, effective Electron app with best practices baked in. Creates a production-ready Electron application with security hardening, modern tooling, proper IPC patterns, auto-updates, native UI elements, and optimal build configuration. Use this skill when users want to start a new Electron project or modernize an existing one.

Synchronise task lists with GitHub Projects using the gh CLI. Use this skill when users want to add tasks, to-dos, or action items to a GitHub repository's project board. The skill can extract tasks from markdown files, user messages, or structured data, then create GitHub issues and add them to projects. It intelligently detects the target repository from git remotes when not explicitly specified.

Generalised autonomous optimisation loop — soft RLVR for any artifact a user can measure. Use this skill whenever a user wants to iteratively improve an artifact — code, prompts, documents, configs, designs, content — by running structured experiments, evaluating results against a multi-dimensional rubric, and learning from each attempt. Triggers include: "optimise this", "keep improving until it's good", "run experiments on", "autoresearch", "iterate on this overnight", "try different approaches and pick the best", or any request implying repeated evaluate-and-improve cycles. Also use when the user wants to improve a system prompt, a data pipeline, a writing style, or any artifact where quality can be decomposed into measurable tracks. For inference optimisation tasks (model latency, throughput, quantization, GPU deployment), a* delegates the low-level tuning to AITune while maintaining quality tracking and learning.

Generalised autonomous optimisation loop — soft RLVR for any artifact a user can measure. Web runtime package: uses memory in this order: connector-backed, project-pack, none. Never assumes subprocess access or unrestricted local files. Use this skill whenever a user wants to iteratively improve an artifact — code, prompts, documents, configs, designs, content — by running structured experiments, evaluating results against a multi-dimensional rubric, and learning from each attempt. Triggers include: "optimise this", "keep improving until it's good", "run experiments on", "autoresearch", "iterate on this overnight", "try different approaches and pick the best", or any request implying repeated evaluate-and-improve cycles.

Creating interactive data visualisations using d3.js. This skill should be used when creating custom charts, graphs, network diagrams, geographic visualisations, or any complex SVG-based data visualisation that requires fine-grained control over visual elements, transitions, or interactions. Use this for bespoke visualisations beyond standard charting libraries, whether in React, Vue, Svelte, vanilla JavaScript, or any other environment.

Use when the user has page annotations from the Missouri Chrome extension, or asks you to look at, highlight, mark up, or comment on elements in a web page. Provides guidance on using Playwright page.evaluate() to read annotations, create markers, focus elements, and interact with Missouri's page API.

Hands-on fine-tuning workshop: guides beginners through training a real LLM on a real dataset using RunPod serverless GPUs for under $20. Use when users ask "how do I fine-tune a model", "I want to try post-training", "train a small model", "fine-tune on RunPod", "QLoRA tutorial", "hands-on fine-tuning", or want to actually train a model rather than just learn theory. Also triggers on "posttrainer", "post-trainer", "chrisvoncsefalvay/posttrainer", or "craft of post-training hands-on". Walks users through account setup, model/dataset selection, launching training, evaluating results and pushing to HuggingFace. For DOING, not reading. Companion to "The Craft of Post-Training" by Chris von Csefalvay (No Starch Press, 2026).