Recipes Search Engine

Recipes-Recommendations is a full-stack recipe search engine powered by custom-trained language models and scalable infrastructure. It’s currently hosted at Recipe-Recommendations.com. There’s also a Github Org with all the source code. The application has 5 main components

  1. The pipeline begins with a Spark-based data processing system that generates over 10 million labeled recipe pairs from raw data.

  2. These pairs are then used to fine-tune a transformer-based model for semantic similarity using PyTorch and Hugging Face.

  3. Once trained, the model generates dense vector embeddings for over 200,000 recipes, which are stored in a Redis Vector Database.

  4. An API takes in user queries, and finds the recipes with the closest vector embedding

  5. React frontend hosted on AWS Amplify provides users with an intuitive interface to search and explore recipe suggestions.

The project showcases the end-to-end development of an AI-powered product—from data engineering and model training to infrastructure and UI. Here’s the tech stack

  • Python and Typescript

  • Apache Spark

  • Kubernetes

  • Docker

  • Argo Workflows

  • Hugging Face Libraries (PEFT, Dataset, Transformers)

  • Python Data Science Libraries (Pandas, Numpy, Scikit-learn)

  • Jupyter Notebook

  • AWS Services (EC2, EKS, S3, ELB, Cloudwatch, Amplify)

  • Redis Vector Database

  • Next.js and React