Agentic AI AUTOMART

🚀 Inspiration

Retail businesses—especially supermarkets—generate massive amounts of data daily, but most store managers lack the tools to turn this data into actionable insights. We were inspired by the gap between data availability and decision-making.

Our goal was to build an intelligent system that not only analyzes data but also thinks like a business assistant, helping managers optimize sales, inventory, and customer engagement in real time.

💡 What it does

Agentic AI AUTOMART is an end-to-end retail analytics platform that combines machine learning models with intelligent agents to provide actionable insights.

It helps supermarket managers:

Understand customer segments (high-value, low-value, etc.) Predict future sales using LSTM models Identify low-stock items and restocking needs Analyze profit and loss at product/store level Discover product bundles using association rules Generate AI-powered business strategies

All of this is delivered through an interactive Streamlit dashboard for easy use.

🛠️ How we built it

We built the system using a modular architecture with three core layers:

Data Layer: Processed CSV datasets using Python libraries like pandas and NumPy ML Models: K-Means → Customer segmentation Apriori → Product recommendations LSTM (PyTorch) → Sales forecasting Agent Layer: Intelligent agents wrap model outputs and convert them into human-readable business insights Frontend: Built using Streamlit to create an interactive dashboard AI Integration: Used Google Gemini via LangChain for generating smart marketing insights ⚠️ Challenges we ran into Integrating multiple ML models into a single pipeline without conflicts Handling dependency issues (like PyTorch, mlxtend, etc.) during deployment Ensuring the system works even without API keys (fallback logic) Designing agents that produce useful business insights instead of raw data Deploying on Streamlit Cloud with CPU-only constraints 🏆 Accomplishments that we're proud of Built a fully functional end-to-end system, not just a prototype Successfully combined ML + AI Agents + Dashboard in one project Created a system that works in real-world retail scenarios Implemented graceful fallback AI logic (works even without Gemini API) Designed a scalable architecture with clean separation of models and agents 📚 What we learned How to design agent-based AI systems instead of traditional pipelines Practical implementation of ML models in business use-cases Importance of data preprocessing and feature engineering Handling real-world issues like deployment, dependency conflicts, and scalability How to convert technical outputs into business-friendly insights 🔮 What's next for Agentic AI AUTOMART Integrate real-time data streaming (IoT / POS systems) Add voice-based AI assistant for managers Improve forecasting with advanced deep learning models Deploy as a SaaS platform for retail chains Add personalized customer recommendation engines Expand to multi-store analytics and supply chain optimization

Built With

• agent
• agent-based
• ai
• ai-powered
• algorithm
• apriori
• architecture
• cloud
• clustering
• control
• csv
• dashboard
• data
• datasets
• deployment
• deployment:
• forecasting
• frameworks
• frameworks:
• gemini
• github
• google
• hosting
• insights
• interactive
• k-means
• langchain
• languages:
• libraries:
• lstm
• matplotlib
• mlxtend
• numerical
• numpy
• operations
• pandas
• platforms
• processing
• python
• pytorch
• sales
• scikit-learn
• storage:
• streamlit
• the
• transactions
• version
• visualization