BrewSmart

Inspiration

Growing up surrounded by small businesses, I've always seen the heart that goes into running an independent coffee shop. But I've also seen what happens when passion meets a lack of information wrong decisions, wasted money, and eventually closed doors.

I want to open a coffee business of my own. Not someday , soon. And when I looked at what separates the coffee shops that thrive from the ones that don't, the answer wasn't the quality of the coffee. It was the quality of the decisions.

Starbucks has an entire data science team. They know exactly what to sell, when to sell it, and what to charge down to the cent. The independent shop owner down the street is guessing. And that gap is killing small businesses.

I'm a Data Science student with a minor in AI. I had the skills to close that gap. So I built BrewSmart.

What it does

BrewSmart is a business intelligence tool built specifically for small coffee shop owners who are already running but flying blind. Upload your sales data a simple CSV or Excel export from your POS system and BrewSmart answers the five questions every coffee business owner needs answered:

• What's actually selling? K-Means clustering groups every product into Stars, Cash Cows, Hidden Gems, and Dead Weight automatically
• When are customers showing up? A peak hours heatmap built from your real transaction timestamps
• Are you leaving money on the table? Price elasticity analysis finds the sweet spot between volume and margin
• Where is the business headed? Linear regression forecasts revenue trends for the next 1–12 weeks
• What should your menu look like right now? A data-backed recommendation built from your top performers

On top of that, BrewSmart includes a Marketing Strategy page with quadrant-based advice for each product group and an Instagram caption generator for your Hidden Gems. And finally, an AI Advisor powered by Claude that reads your entire dataset and answers business questions in plain language like having a consultant who actually knows your numbers.

How we built it

BrewSmart is built entirely in Python using the following stack:

• Streamlit — web app framework, turns Python directly into a browser interface
• pandas + NumPy — data loading, cleaning, and transformation
• scikit-learn — K-Means clustering with StandardScaler normalization
• Plotly — all interactive charts and visualizations
• Linear Regression — implemented via NumPy's polyfit for revenue forecasting
• Anthropic API (claude-opus-4-6) — powers the AI Advisor chat feature

Challenges we ran into

Data flexibility was the hardest problem. Every POS system exports data differently. Building a column detection system that could recognize transaction_qty, qty, units_sold, and count as the same thing without breaking took significant iteration.

Accomplishments that we're proud of

Built a fully working, dataset-agnostic ML-powered web app in under 24 hours

• The K-Means clustering model correctly identifies product performance categories across completely different datasets without any manual tuning
• The AI Advisor genuinely answers specific business questions grounded in real data not generic advice
• The app works with the Kaggle dataset out of the box but is flexible enough to accept any coffee shop's real POS export
• The Marketing Strategy page generates actual ready-to-post Instagram captions a feature that surprised us with how useful it felt in practice

What we learned

• Data science is most powerful when it solves a personal problem. Every decision about what to build was easier because I understood the user because I am the user.

Streamlit is incredibly powerful for data science products. Going from Python analysis to a fully interactive web app in the same language is a superpower for data scientists who aren't primarily engineers.

What's next for BrewSmart

The vision is bigger than one coffee shop.

Short term:

• Connect directly to Square and Toast POS APIs so owners don't need to export manually data flows in automatically
• Add inventory tracking so the model can factor in cost of goods and calculate true profit margin per product, not just revenue

Medium term:

• Expand beyond coffee to any small food and beverage business — every bodega, food truck, and pop-up has this exact same problem
• Build a multi-location comparison feature for owners with more than one spot

Long term:

• A platform where any small business owner uploads their data and gets instant intelligence no data scientist required, no consultant fee, no MBA needed
• The same tools that big chains use, available to the independent owner who deserves them just as much

Small businesses are the backbone of our communities. They just need better tools. BrewSmart is the first step toward giving them that.

Built With

• claude
• cursor
• machine-learning
• python
• streamlit