Hugging Face plugins

Build and deploy production-grade LLM applications with LangGraph for agent orchestration, advanced RAG pipelines leveraging vector and hybrid search, prompt engineering patterns, and automated evaluation. Covers embedding model selection, vector index optimization, and multi-agent architectures for document Q&A, chatbots, and semantic search over proprietary data.

Manage the full Hugging Face ML lifecycle from a single agent: search and select models, estimate GPU memory, train or fine-tune with TRL/Unsloth, evaluate locally, build and deploy Gradio demos on Spaces, publish datasets and research papers, and run models in-browser with Transformers.js.

Explain machine learning model predictions using SHAP, LIME, and feature importance to identify influential features and debug behavior. Generate production-ready AI/ML code from context, including validation, error handling, performance metrics, insights, artifacts, and documentation.

Fine-tune pre-trained ML models like ResNet, BERT, and GPT on custom datasets using transfer learning. Generates production-ready Python code with validation, error handling, performance metrics, documentation, and saves artifacts for deployment.

Optimize Python deep learning models using Adam, SGD optimizers, learning rate schedulers, and regularization to improve accuracy and reduce training time. Generate production-ready AI/ML code from context analysis, including validation, error handling, performance metrics, insights, artifacts, and documentation.

Build recommendation engines by generating Python code for collaborative, content-based, or hybrid filtering using scikit-learn, TensorFlow, or PyTorch to personalize movies, products, or content. Analyze context to produce complete AI/ML tasks with validation, error handling, performance metrics, insights, artifacts, and documentation.

Deploy, debug, optimize, monitor, and secure GPU-accelerated ML inference and training workloads on CoreWeave Kubernetes clusters, including cost tuning, data handling, migrations from AWS/GCP/Azure, CI/CD automation, and production checklists.

Evaluate single ML models or compare multiple ones on test datasets across classification, regression, NLP, and generative tasks. Compute metrics, statistical significance, inference performance, costs, robustness, bias checks; generate visualized reports with confusion matrices, performance profiles, tables, rankings, and recommendations.

Build, train, and deploy AI models on Amazon SageMaker — validate datasets, select fine-tuning techniques, run SFT/DPO/RLVR training, diagnose HyperPod cluster issues (NCCL, GPU, Slurm), and deploy to endpoints or Bedrock, all from your coding assistant.

Delegate image analysis, OCR text extraction, barcode/QR detection, and document processing to a vision expert agent using latest models like GPT-4V, Claude Vision, Mistral-OCR, Tesseract, and EasyOCR for efficient visual AI workflows.

Rapidly implement production-ready AI/ML features in apps: integrate LLMs with prompt engineering and response handling, build ML pipelines for recommendation systems, add computer vision for visual search, and enable intelligent automation using OpenAI, Anthropic, LangChain, Hugging Face, or Ollama.

Store, search, and recall persistent cross-session memories for Claude Code using Ebbinghaus decay, enabling the AI to retain user facts, preferences, and project decisions across conversations.

Rapidly implement production-ready AI/ML features in apps: integrate LLMs via prompt engineering and response handling, build ML pipelines for user behavior-based recommendations, add computer vision for photo-based product search, and deploy intelligent automations.

Rapidly implement production-ready AI/ML features in apps, including LLM integrations with prompt engineering, ML pipelines for recommendations, computer vision for visual search, and intelligent automation, using a specialized agent.

Build RAG pipelines for document Q&A and chatbots by chunking large docs, generating embeddings, storing in vector DBs, and retrieving context to reduce hallucinations. Engineer and optimize LLM prompts using chain-of-thought, few-shot examples, constitutional AI, meta-prompting, and validation workflows.

Orchestrate medical research projects end-to-end: intake and scaffold manuscripts, run systematic reviews and meta-analyses, audit citations and reporting guideline compliance, generate publication-ready figures and statistical code, de-identify clinical data, and produce IRB protocols, peer reviews, and journal submission packages.

Delegate advanced image analysis workflows to expert vision AI subagents that perform OCR with Tesseract/EasyOCR, barcode/QR detection, document processing, and optimization using cutting-edge models like GPT-4V, Claude Vision, and Mistral-OCR.

Enforces feature-layer architecture for Claude Code projects with battle-tested skills: scaffold bounded layers and feature narratives, run multi-agent swarm reviews, freeze acceptance criteria, and generate pixel art. Includes iOS development, Remotion video production, and humanization tools for AI-generated text.

Migrates AI models and custom operators to Huawei Ascend NPUs, handles GPU-to-NPU code adaptation, profiling, performance optimization, distributed training deployment, and provides developer tooling for vLLM inference, Ascend C/Triton kernel development, and environment diagnostics.

Delegate end-to-end ML engineering workflows to specialized agents that construct data preparation and training pipelines with feature engineering and hyperparameter tuning, optimize inference through quantization, pruning, batching, and edge deployment, and manage MLOps for model versioning, monitoring, A/B testing, and production orchestration.

Consult a virtual CTO team of specialized agents to design scalable architectures, generate phased roadmaps with effort estimates, recommend tech stacks, validate plans with ruthless reports identifying flaws and risks, challenge assumptions, estimate costs, and decide build-vs-buy for web, mobile, or AI/ML projects.

Deploy vLLM OpenAI-compatible inference servers locally with hardware detection, via Docker images, or Kubernetes YAML manifests with GPU support, then benchmark throughput, TTFT, TPOT, inter-token latency, and prefix caching using synthetic data, ShareGPT, or fixed prompts.

Turn Claude Code into a proactive engineering partner with 24 skills across 5 layers, enforcing TDD, root-cause debugging, domain-specific patterns for ML/embedded/AI, multi-agent swarm execution, context window management, git worktree isolation, and model-tier routing to cut API costs by 50-65%.

Diagnose and fix ML training failures (OOM, NaN, divergence), generate citation-grounded implementation plans for fine-tuning and inference pipelines, and verify code/configs against official framework docs before running GPU jobs.

Run GGUF models locally with Mozilla Llamafile, launching OpenAI-compatible API servers configurable for GPU/CPU inference, SDK integrations, installations, startups, and connection troubleshooting in offline setups.

Build high-quality computer vision datasets and models with FiftyOne: import/export datasets, run inference, evaluate predictions, find duplicates, visualize embeddings, and develop custom plugins — all from within Claude.

Find, compare, run, and prompt AI models hosted on Replicate directly from the editor, including building custom models with Cog, searching by task, and deploying via CI/CD

Generate, edit, and animate images and videos directly in your terminal by building and executing ComfyUI Stable Diffusion workflows, with support for advanced models like Flux, WAN, Qwen, and LTX-V2

Routes OpenAI-compatible clients through FreeRide, a local gateway that distributes inference requests across free-tier AI providers (OpenRouter, Groq, NVIDIA NIM, Cloudflare Workers AI, HuggingFace). Detects a running FreeRide instance and wires any OpenAI-shaped client against it for cost-free model access.

Automate end-to-end ML performance investigations: research SOTA papers and architectures, generate phased plans, judge experimental methodologies, profile bottlenecks, run metric-improvement campaigns with atomic git commits, auto-rollback on regressions, and leverage specialist agents for data lifecycle and deep paper analysis.

Enforce a rigorous empirical research pipeline for ML/AI claims: extract competitor baselines, preregister hypotheses, run adversarial falsification, execute locked experiments with statistical checks, and force kill-or-ship decisions based on repository evidence. Integrates with Hugging Face Hub for model training, evaluation, dataset inspection, and paper retrieval.

Automate image processing, OCR, barcode/QR detection, and document analysis with expert vision AI. Proactively select optimal models like GPT-4V, Claude Vision, or Mistral-OCR; engineer prompts, preprocess images, benchmark performance, and integrate APIs from OpenAI, Anthropic, and Hugging Face.

Build full-stack dApps on Ritual Chain using skills and agents that generate Solidity smart contracts with async precompiles for on-chain ML inference, HTTP/LLM calls, scheduling, secrets, X402 payments; create React/Next.js frontends with wagmi hooks; set up backends, testing, debugging, and deployment workflows from idea to verification.

Build production-grade LLM apps in Python: implement RAG pipelines with embeddings and hybrid search, design LangChain/LangGraph agents, optimize prompts, tune vector indexes, and evaluate performance using AI agents, skills, and commands for architecture, code gen, and benchmarking.

Transcribe handwritten historical Swedish documents using HTR tools with interactive viewers and JSON exports, search Riksarkivet archives via metadata and transcriptions, browse and zoom PDFs/IIIF images, annotate pages in Label Studio, map topics to record types, and upload files to Hugging Face Spaces for remote access.

Run semantic (vector) and full-text (keyword) searches across indexed PDFs, markdown, and source code using the arc CLI. Index content into Qdrant and MeiliSearch with AST-aware chunking, frontmatter, and git metadata, then query either search type via a single command.

Reproduce Long Video Sparse Attention (LVSA) paper headline numbers using bundled benchmarks scripts, including SotA comparison, latency scaling, and scoring with VQeval and VBench-Long, plus figure regeneration.

Delegate expert-level AI/ML workflows to specialized agents: engineer optimized prompts with evaluation and A/B testing, architect scalable LLM systems with RAG/LoRA fine-tuning, build production NLP pipelines for NER/classification/QA, and deploy optimized models via vLLM/Triton/Docker/K8s for reliability, performance, and cost control.

Train, evaluate, export, and deploy NVIDIA TAO models for computer vision tasks including classification, detection, segmentation, pose estimation, depth estimation, OCR, and video analysis. Orchestrate AutoML hyperparameter optimization, DEFT iterative data improvement loops, and synthetic data generation. Submit GPU-accelerated training jobs across Docker, Kubernetes, SLURM, Brev, and DGX Cloud platforms.

Build dynamic neural networks that grow, prune, and adapt topology during training in Python with Hugging Face. Design state machines for module lifecycles, diagnose issues in growth decisions, gradient isolation, and integration, plus consult an expert advisor for continual learning, PEFT/LoRA, and modular composition.

Train and run inference on machine learning models using Hugging Face Transformers and PEFT with PyTorch on cloud GPUs from Modal, Lambda Labs, or RunPod—no local GPU required.

Discover, evaluate (quality, licensing, provenance), and acquire datasets from Kaggle, HuggingFace, IPFS, arXiv, DBLP for AI model training and fine-tuning. Download free or paid data via local stdio MCP server—no API keys needed.

Build and debug robotic dataflow applications with dora-rs: configure YAML dataflows, develop Rust/Python nodes, integrate ML/vision/audio pipelines, control robots, and record demonstrations for imitation learning.