Científicos de datos

The-Transition-Accelerator0

Edit Archetype

Edit or create an archetype CSV file in an Energy Wallet data product. Guides analysts through population share distributions with structural validation, downstream impact analysis, and data integrity enforcement.

LLM & AI

Deep Research

Conduct deep multi-source research on any topic. Use when the user wants comprehensive research with citations, multiple perspectives, and synthesis. Triggers: "research X", "deep dive on X", "comprehensive analysis of X", "investigate X thoroughly", "find out everything about X". Outputs to Obsidian vault with chat summary.

thewelshrich0

Contratos Inteligentes

Data Audit

Audit complet de qualite data sur les 9 dimensions ISO 8000-8. Utiliser pour verifier la fiabilite des donnees d'un projet, d'une table, ou d'un SIREN specifique.

LLM & AI

Company Name Fuzzy Matching

Multi-strategy matching with length-aware thresholds and stopword filtering

RevPalSFDC0

Dl Scoring

Scoring v3 hotLeadScore (5 axes, poids, signaux enrichment)

Lead Scorer Ml

Scoring ML hybride des prospects : LLM au démarrage, RandomForest quand assez de données. Utiliser pour évaluer la probabilité de conversion d'un prospect.

Data Cleaner Agent — Nettoyage & Structuration

Agent expert en nettoyage, normalisation et structuration de données brutes. Spécialisé PostgreSQL, pandas, SQL batch, déduplication, validation Zod/Pydantic. Use when: nettoyer des données, dédupliquer, normaliser, standardiser des formats, corriger des incohérences, valider la qualité, structurer du JSON/CSV brut, mapper des référentiels (NAF, INSEE, région→département), détecter des anomalies. Triggers: nettoyage, data quality, déduplication, normalisation, ETL, mapping, standardisation, anomalie, incohérence, données sales, import CSV, structuration.

retinal-information-processing-lab0

Which Price Forecast Model

States which model powers the in-app near-term stock price outlook (AI Research page) versus which pretrained Amazon/Hugging Face model is only documented. Use when the user asks which AI or model is used for price prediction, forecast, Chronos, Amazon, Holt, /which, or confuses the research LLM (OpenRouter) with the price outlook block.

revanth-rampal0

Química Computacional

Gradients

Load analytical gradient formulas for arc-cosine kernel

Jupyter Ml

Full CUDA ML JupyterLab image with real-time collaboration and CRDT MCP server. Base: nvidia. Port 8888. GPU-accelerated ML training + collaborative notebooks. MUST be invoked before building, deploying, or troubleshooting the jupyter-ml image.

overthinkos0

Motion Recognition ML Pipeline

End-to-end ML pipeline for motion/gesture recognition using the WRLDS AirHive 9-axis IMU sensor (accelerometer + gyroscope). Use when building a movement classifier from BLE sensor data — covers data collection format, labeling, feature extraction, model training, evaluation, and deployment. Triggers on requests like: train a model to recognize gestures, classify sensor movements, detect specific motions, build a sports analytics ML model, or set up a motion recognition pipeline for a customer project.

wrlds-creations0

MacroMate PyTorch Model Integration Skill

MacroMate new PyTorch model integration skill. Use this when replacing the old final_model.h5 (Keras/TensorFlow) with the new best_model.pt (PyTorch EfficientNet) trained on Food-101 + Indian foods + UECFood256. Covers: placing the 3 new files, completely rewriting ml_food_predictor.py to load best_model.pt with torchvision, hybrid detection logic (custom model first → CLIP fallback), nutrition lookup from label_nutrition_mapping.json, showing model name + confidence badge to the user in result.html and multi_result.html, and updating nutrition_provider.py. Apply alongside macromate-ui skill for visual consistency.

renish76060

Python Ml

GPU-accelerated Python ML environment with CUDA, PyTorch, and llama.cpp. No Jupyter server — use as a base for ML workloads or interactive shell. MUST be invoked before building, deploying, configuring, or troubleshooting the python-ml image.

overthinkos0

Contenedores

Notebook Finetuning

Unsloth fine-tuning notebook collection provisioned into the workspace volume at deploy time. Data-only layer — no packages, no services, no dependencies. Use when working with notebook-finetuning, Unsloth training notebooks, or unsloth-studio data provisioning.

overthinkos0

LLM & AI

autoresearch

Autonomous ML experimentation framework by Andrej Karpathy. AI agent autonomously modifies train.py, runs 5-minute GPU experiments, evaluates with val_bpb, and commits only improvements via git ratcheting — so you wake up to 100+ experiments and a better model. Use when setting up autoresearch, writing program.md directives, interpreting results, configuring hardware, or running overnight autonomous ML experiments. Triggers on: autoresearch, autonomous ml experiments, overnight gpu experiments, karpathy autoresearch, train.py experiments, val_bpb, program.md research directives, ai runs experiments.

Reinasboo0

Académico

Mnemos Research

Conduct research for mnemos using current technical sources, prior art, and literature-backed analysis, then write a focused research note or update hypotheses. Use when investigating papers, similar repositories, benchmarks, current best practices, or design trade-offs.

reeinharddd0

Model Export

Export and quantize ML models for deployment. Use when the user says "export model", "GGUF", "llama.cpp", "quantize", "ONNX", "TorchScript", "deploy model", "convert model", "run locally", "Ollama", or "inference server".

Hf Finetune

Fine-tune a HuggingFace transformer model with Trainer API or custom loop. Use when the user says "fine-tune", "finetune", "HuggingFace", "Trainer", "SFT", "PEFT", "LoRA", "custom LLM", "adapt a model", or "train a transformer".

Training Infra

Set up production training infrastructure with distributed training, experiment tracking, checkpointing, and reproducibility. Use when the user says "multi-GPU", "distributed training", "W&B", "Weights and Biases", "MLflow", "DDP", "experiment tracking", "resume training", or "reproducible".

Unsloth Finetune

Fine-tune LLMs with Unsloth for 2x+ faster training and lower memory usage. Use when the user says "Unsloth", "unsloth", "fast fine-tuning", "efficient LoRA", or wants to fine-tune Llama, Mistral, Qwen, or Gemma efficiently.

Pytorch Lightning

High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.

overviewlabs0

Astronomía y Física

Timedatamodel

Guides creation and manipulation of TimeDataModel objects (TimeSeries, TimeSeriesTable). Use when working with the timedatamodel Python library, time series with rich metadata, DataShape, Frequency, DataType, GeoLocation, or energy/forecasting data pipelines.

rebase-energy0

Rh Inf Extract

Reviewer-gated extraction skill for deriving L2 structured artifacts from ingested normalized sources. Modes: plan · implement · verify.

reason-healthcare0

Huggingface Accelerate

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

overviewlabs0

Schema Query

Query the CDM DuckDB database using LinkML schema awareness. Use this skill when the user wants to query the database with understanding of the data model, relationships, and semantics defined in the LinkML schema. This provides richer, more intelligent queries than basic SQL.

realmarcin0

Documentos

Data Analyst

Expertise in analyzing CSV and JSON data files. Use when the user asks to "summarize", "analyze", or "filter" a data file in the workspace.

realfakenerd0

Purpose

Use when the user wants to profile a PyTorch program with torch.profiler, identify CPU/CUDA hotspots, inspect kernel or stream timing, understand CUDA Graph behavior, profile distributed or NCCL overlap, or when they mention "torch.profiler", "PyTorch profiler", "profiler trace", "key_averages", "export_chrome_trace", "hotspot", "kernel timing", "dataloader overhead", "profiler schedule", "TensorBoard profiler", or "profile step".

rdspring10

Seguridad

Approve Data Quality Specification

Approves a Data Quality Specification (DQS) document by setting its status to Approved. Reads the latest DQS, verifies it is not already approved, updates the Status field and adds a version history entry. This is the gate that enables downstream artifact creation (LLD, Stories). Use when the user asks to: - Approve a DQS - Sign off on a DQS - Mark a DQS as approved or ready for handoff

RDEWAI0