All mini projects 
Neural-network-driven cars evolving racing behavior
> cat ~/artifacts/nn-racer/README.md
NN-Racer
CompletedNeural network evolution using genetic algorithms for autonomous racing behavior.
Date
2024-05
Duration
2 months
Team
solo
Difficulty
hard
Project Story
NN-Racer is an experiment in evolutionary learning where neural networks control racing agents and improve over generations.
The project shows how mutation and selection can discover performant control policies without classic backpropagation loops.
Technical Details
Tech Stack
Key Features
Evolutionary training loop
Fitness-based selection
Autonomous behavior discovery
Generation-by-generation visualization
Multi-objective optimization
Challenges Faced
High compute cost for many generations
Hyperparameter sensitivity
Local optima traps
Exploration versus exploitation balance
Key Learnings
Fitness design controls learning direction
Evolution discovers unexpected strategies
Architecture choice changes convergence dynamics
Long-running experiments need strong observability
Need a similar implementation?
If you want to build a practical AI feature like this in your product, reach out — I can help with architecture, prototyping, and delivery.