agentleFS
Sign inStart free
TheWatcher01TheWatcher01/skills/.claude/skills/data-science/SKILL.md

data-science

Skill0 starsChanged 7 months agoMITView on GitHub

Agent data scientist spécialisé dans l'exploitation du datalake SubventionAI (2.8M entités, 168K versements, scoring ML, embeddings bge-m3 1024d). Stack: Python, PostgreSQL+pgvector, scikit-learn, XGBoost, LightGBM, spaCy, pandas. Use when: feature engineering, model training/evaluation, scoring batch, analyse statistique sectorielle, corrélation, clustering, prédiction, classification, anomaly detection, NLP sur objets sociaux, requêtes SQL analytiques, optimisation hyperparamètres. Triggers: data science, ML, modèle, scoring, feature engineering, prédiction, classification, clustering, analyse sectorielle, fundability, anomalie, NLP, embeddings, train, evaluate, cross-validation, AUC, F1, SHAP, Optuna.

Create a free account to mount
Files
SKILL.md

Data Science Agent — Datalake SubventionAI

Données disponibles

Consulter references/schema.md pour le schéma complet.

TableRowsClé
dl_entities2.8MSIREN, NAF, géo, scores, industryData JSONB
dl_subventions_versees168KHistorique versements réels 2010-2026
dl_subventions2413Programmes actifs
dl_subvention_anomalies3.1KAnomalies détectées
dl_bodacc_annonces-Annonces légales
dl_marches_publics-Marchés publics

Connexion : PGPASSWORD=datalake_dev psql -h 127.0.0.1 -p 5434 -U datalake -d datalake

Workflow ML

  1. Extraction : SQL paramétré, export CSV pour batch compute
  2. Feature engineering : pandas + NumPy dans scripts/ml/
  3. Training : scikit-learn/XGBoost/LightGBM, cross-validation 5-fold
  4. Evaluation : AUC-ROC, F1, precision/recall, SHAP feature importance
  5. Tuning : Optuna (TPE sampler, 100 trials)
  6. Scoring batch : export → compute local → reimport PG (séparer compute et I/O)
  7. Persistance : colonnes dédiées dans dl_entities

Scoring existant

ScoreColonneModèle
FundabilityfundabilityScoreBatch SQL 6 critères
Digital maturitydigitalMaturityScore7 facteurs rule-based
CNIL riskcnilRiskScoreXGBoost AUC 0.90
Hot leadhotLeadScoreComposite 6 dimensions

Règles critiques

  • SQL toujours paramétré ($1, $2) — jamais d'interpolation
  • Données factuelles uniquement — jamais de chiffres approximatifs
  • Performance first : batch SQL, COPY pour imports, pas de N+1
  • Scripts dans scripts/ml/ avec docstring et if __name__ == "__main__":
  • Modèles évalués avec cross-validation avant déploiement
  • Embeddings : bge-m3 1024d, colonne vector(1024) pgvector, index HNSW

Patterns courants

Consulter references/patterns.md pour templates SQL et Python.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free