Automating XGBoost Hyperparameter Tuning with Agentic AI
summary
Presents an implemented multi‑agent system that automates XGBoost hyperparameter search with five agents (parameter generator, trainer, evaluator, tracker, selector), replacing grid/random search and Excel logging. Runs 247–500+ automatic configurations, cuts manual tuning from 5–7 days → 2 hours, and logs every run for traceability. Reports uplift from 0.84 → 0.91 F1 (+7%) and accuracy from 88% to 94%, claims ~95% time saved and continuous retuning for drift.
AI-written summary · about
chapters (14)
- 0:00Agentic AI for XGBoost tuning
- 1:26Manual tuning fails
- 5:41XGBoost optimization challenges
- 8:35Agentic AI replaces manual search
- 11:02Multi-agent XGBoost architecture
- 13:13Tracking agent decisions
- 15:22Time saved vs manual tuning
- 17:53Accuracy and F1 gains
- 19:17Extending to deep learning
- 21:49Fairness transparency cost
- 23:14Future-proofing models
- 24:42Agentic AI next steps
- 26:21Beyond XGBoost
- 27:02Q&A on agentic AI
related talks
26mModel Selection with Python: An Introduction to Hyper Parameter Tuning
01h 21mAgent Optimization with Pydantic AI: GEPA, Evals, Feedback Loops
18mAlexandra Johnson, Software Engineer, SigOpt, NYC 2017
25mFrontiers of Agentic AI
36mTowards automating machine learning
38mA Guide to Modern Hyperparameters Turning Algorithms
41mOptimizing Machine Learning Models
33mAutomatic Agent Optimization and the Rise of Self-Improving Agents