ML Training Pipeline

advanced · 4 agents · 9 skills

An end-to-end machine learning pipeline spanning four agents. Data preparation, model training, evaluation, and serving each run on dedicated hardware. Models and datasets flow over encrypted Pilot tunnels with automatic approval gating before production serving.

Install

clawhub install pilot-ml-training-pipeline-setup

Skills used

Agents

<your-prefix>-data-prep - Data Preparation
Cleans, validates, and transforms raw datasets. Shares processed data with the trainer.
Skills: pilot-dataset, pilot-share, pilot-task-chain
<your-prefix>-trainer - Model Trainer
Receives prepared datasets, runs training jobs, tracks metrics, and shares trained model artifacts.
Skills: pilot-dataset, pilot-model-share, pilot-metrics, pilot-task-chain
<your-prefix>-evaluator - Model Evaluator
Scores trained models against benchmarks and gates promotion to serving.
Skills: pilot-model-share, pilot-metrics, pilot-review, pilot-task-chain
<your-prefix>-serving - Model Server
Loads approved models, serves inference, monitors health, and load-balances.
Skills: pilot-model-share, pilot-health, pilot-webhook-bridge, pilot-load-balancer, pilot-metrics

Data flows

<your-prefix>-data-prep → <your-prefix>-trainer:1001 - cleaned datasets
<your-prefix>-trainer → <your-prefix>-evaluator:1001 - model checkpoints and metrics
<your-prefix>-evaluator → <your-prefix>-serving:1001 - approved models
<your-prefix>-serving → <your-prefix>-evaluator:1002 - inference metrics for drift detection

Quick start

# Replace <your-prefix> with a unique name for your deployment (e.g. acme)
# On data processing node
clawhub install pilot-dataset pilot-share pilot-task-chain
pilotctl set-hostname <your-prefix>-data-prep

# On GPU training node
clawhub install pilot-dataset pilot-model-share pilot-metrics pilot-task-chain
pilotctl set-hostname <your-prefix>-trainer

# On evaluation node
clawhub install pilot-model-share pilot-metrics pilot-review pilot-task-chain
pilotctl set-hostname <your-prefix>-evaluator

# On serving node
clawhub install pilot-model-share pilot-health pilot-webhook-bridge pilot-load-balancer
pilotctl set-hostname <your-prefix>-serving
# On data-prep:
pilotctl handshake <your-prefix>-trainer "setup: ml-training-pipeline"
# On trainer:
pilotctl handshake <your-prefix>-data-prep "setup: ml-training-pipeline"
# On evaluator:
pilotctl handshake <your-prefix>-serving "setup: ml-training-pipeline"
# On serving:
pilotctl handshake <your-prefix>-evaluator "setup: ml-training-pipeline"
# On evaluator:
pilotctl handshake <your-prefix>-trainer "setup: ml-training-pipeline"
# On trainer:
pilotctl handshake <your-prefix>-evaluator "setup: ml-training-pipeline"
pilotctl trust