Slice Price

# Cut prices, not corners.

Comment

Inspiration

We wanted to simplify the lives of sauce bros — the legends running pizza joints — by impressing them with smart analytics and leveraging emerging technologies, all wrapped in a sleek and modern UI.

What It Does

Slice Price offers three powerful features:

  1. Sales Predictions: Forecast future pizza sales to help restaurant owners plan ahead.
  2. Market Comparison: Analyze how your restaurant stacks up against nearby competitors and pizza places.
  3. Inventory Tracking: Use computer vision to automatically scan and count inventory from images — no manual work needed.

How We Built It

  • For predictions, we started by building a custom neural network using TensorFlow.
  • To go beyond basic pricing or trend data, we integrated Prophet, a time-series forecasting library by Meta, to make more accurate and scalable sales forecasts.
  • For inventory tracking, we built a custom OpenCV Transformer model, trained on a dataset we created ourselves. The model analyzes images and counts the number of items accurately.
  • All of this is packaged into a clean UI using Reflex.

Challenges We Ran Into

While the backend was fairly smooth, the front-end UI gave us a hard time. Styling bugs, layout issues, and visual glitches made debugging very tedious. Making the interface look sleek and perform well took a lot of effort.

Accomplishments That We're Proud Of

Even though we were just a two-person team, we managed to complete the entire application within a 24-hour window. Everything came together nicely, looks great, and is easy to use — which we’re super proud of.

What We Learned

  • Learned a lot about time-series forecasting and integrating tools like Prophet into a real-world application.
  • It was our first time using Reflex for the frontend, and though it was challenging, we pushed through and got it done.
  • We also gained experience in building custom datasets and training CV models from scratch.

What's Next for Slice Price

  • Add an inventory checklist view, where the user can take a photo of their storage room and get an automatic analysis of whether their stock is sufficient.
  • Introduce a new input tab that allows users to feed more custom data into the system — improving the precision of predictions and providing deeper insights.
  • Optimize the computer vision model to handle more complex environments and varied inventory layouts.
  • Possibly roll out beta testing with local pizza spots to gather real-world feedback.

Built With

Share this project:

Updates