Supply chain solution

• 

  Forecast demand with plots of confidence intervals and predicted demand.

• 

  Sample forecast containing upper and lower bound of the prediction interval and the forecast for current time.

Inspiration

Supply chain disruptions cost companies billions annually. Forecasting demand accurately and making accurate and quick inventory decisions is crucial. I wanted to build a GenAI-powered tool that helps businesses forecast demand and optimize inventory using real-world data, making operations leaner, faster, and smarter.

What it does

-Upload retail data -Generate demand forecasts using time series models (e.g., Prophet) -Visualize demand trends with confidence intervals -Cluster products based on demand behavior -Receive inventory recommendations -Optionally upload/output files to AWS S3

How we built it

Frontend: Streamlit for fast prototyping and UI Backend: Python for data preprocessing Prophet for forecasting KMeans+DBSCAN+PCA algorithms for clustering Storage: AWS S3 to store forecast output files Compute: AWS EC2 (Ubuntu instance) for running the app and scripts Data: UCI Online Retail dataset (cleaned)

Challenges we ran into

Handling a large number of CSVs and slow I/O due to EC2 specs AWS credential management (permissions, access keys) Issues with Streamlit deployment and networking (public IP vs local IP) External package installs in a restricted environment App performance tuning and keeping it within the free tier limits

Accomplishments that we're proud of

Fully deployed a production-ready forecasting dashboard using cloud infrastructure Built a robust forecasting and inventory recommendation system Used real-world data to simulate business use cases Overcame AWS, environment, and networking hurdles to ship a live demo

What we learned

How to work with AWS EC2 and S3 Best practices for secure credential management Handling time series forecasting at scale Combining machine learning, cloud, and web development The importance of keeping things lightweight for free-tier compliance

What's next for Supply chain solution

Add authentication and multi-user support Enable real-time data streaming from POS systems Introduce automated restocking suggestions via APIs Explore integration with AWS Lambda and DynamoDB Improve UI and add filters/search to explore forecasts interactively

Built With

• amazon-ec2
• amazon-web-services
• anaconda
• boto3
• commandline
• github
• iam
• matplotlib
• numpy
• pandas
• plotly
• prophet
• python
• s3
• scikit-learn
• scp
• ssh
• streamlit
• ubuntu
• vscode