Built a simulation engine to optimize warehouse picking efficiency by testing SKU-slotting strategies before real-world implementation.
In simulations, dynamic SKU placement based on real-time order frequency reduced average pick-path distance by ~15% compared to static storage methods.
Warehouse operations often rely on static SKU placement, leading to inefficient picking routes and increased operational time.
Testing new layouts in real environments is costly and risky.
Developed a configurable simulation system that models warehouse operations and allows operators to experiment with different inventory strategies.
The system enables:
- Comparison of SKU-slotting policies
- Visualization of high-traffic zones
- Rapid iteration without operational disruption
-
Simulation Engine
Models picking workflows and SKU-slotting strategies using Python -
Policy Switching Dashboard
Compare frequency-based vs profit-margin-based placement in real time -
Spatial Heatmaps
Identify warehouse “hot zones” for optimization -
Data Layer
Supabase (PostgreSQL) backend handling high-concurrency logs -
Configurable Architecture
YAML-based system for flexible warehouse/environment setup
Orders → Simulation Engine → Policy Logic → Path Calculation → Output Metrics + Heatmaps
- Python
- Streamlit
- Supabase (PostgreSQL)
- YAML
Dynamic SKU-slotting based on demand frequency significantly reduces picker travel distance, improving operational efficiency without requiring infrastructure changes.
For a production-grade deployment:
- Introduce workflow orchestration (e.g., scheduled runs, retries)
- Integrate real-time warehouse data streams
- Add monitoring for simulation accuracy vs real-world outcomes
-
Clone the repository
-
Install dependencies:
pip install -r requirements.txt
-
Configure your Supabase credentials in the
.envor YAML configuration layer -
Run the simulator:
streamlit run app.py