Overview
SerpentLab is a hybrid Go and Python project built around reinforcement learning experiments and simulation design.
SerpentLab Architecture
SerpentLab separates simulation logic, training workflows, and orchestration so experiments can evolve without tightly coupling the system.
sim
Simulation Engine
Runs game state and environment transitions.
trainer
Training Loop
Consumes environment signals and updates model behavior.
controller
Experiment Controller
Coordinates training runs and configuration.
metrics
Metrics / Evaluation
Tracks reward, performance, and run outcomes.
config
Config Layer
Defines experiment settings and run parameters.
Service / Data Flow
Config Layer→Experiment Controller
Experiment configurationExperiment Controller→Simulation Engine
Starts simulation runsSimulation Engine→Training Loop
State / reward signalsTraining Loop→Simulation Engine
Action decisionsTraining Loop→Metrics / Evaluation
Training resultsExperiment Controller→Metrics / Evaluation
Run metadataWhat I built
- a simulation-oriented environment for experimentation
- separation between simulation logic and training logic
- service-friendly boundaries that make the project easier to evolve
- a playground for testing ideas around AI workflows and systems design