SmartShop: Personalized Grocery Item Recommendation System

January 2023 – May 2023

Machine Learning Deep Learning Python TensorFlow FastAPI

Overview

SmartShop is an intelligent recommendation system that provides personalized grocery item suggestions based on user preferences, purchase history, and behavioral patterns. This project resulted in a peer-reviewed publication at ICONIC2023 conference.

Key Features

  • Personalized Recommendations: ML-powered recommendation engine using collaborative and content-based filtering techniques.
  • User Profiling: Advanced user profiling system that learns from purchase patterns and preferences.
  • Real-time Predictions: Fast, real-time recommendation generation through optimized model serving.
  • Visual Product Search: Computer vision-based product identification for enhanced user experience.
  • Seasonal Trends Analysis: Incorporation of seasonal buying patterns and trending products.
  • A/B Testing Framework: Built-in experimentation framework for model performance comparison.

Technical Implementation

  • Data Processing: Pre-processed and curated large-scale grocery transaction datasets with 100K+ products.
  • Feature Engineering: Engineered advanced features including user demographics, temporal patterns, and product attributes.
  • Model Development: Benchmarked multiple architectures:
    • Collaborative filtering using matrix factorization (SVD, NMF)
    • Deep neural networks with embedding layers
    • Hybrid models combining content and collaborative approaches
    • Sequence models (LSTM, GRU) for temporal pattern recognition
  • Visualization: Used t-SNE and UMAP for embedding visualization and cluster analysis.
  • API Development: Deployed scalable RESTful API using FastAPI with async processing.
  • Team Leadership: Led a cross-functional team of 4 engineers, managing sprint planning and code reviews.

Results & Impact

  • Achieved 78% recommendation accuracy on test dataset.
  • Improved user engagement by 35% through personalized suggestions.
  • Reduced product search time by 45% with intelligent recommendations.
  • Successfully deployed to production serving 10K+ daily active users.
  • Published peer-reviewed research paper at ICONIC2023 conference.

Technologies Used

ML/DL: TensorFlow, PyTorch, Scikit-learn, Surprise
Data Processing: Pandas, NumPy, PySpark
Visualization: Matplotlib, Seaborn, Plotly, t-SNE
API: FastAPI, Redis (caching)
Database: PostgreSQL, MongoDB
Deployment: Docker, AWS (EC2, S3)

Publication

SmartShop: Personalized Grocery Item Recommendation System
Presented at ICONIC2023
Authors: Yash Kumar, Hasib Ahmed, Himanshu Kumar Sinha

← Back to Projects