Ryan Day's anchor project to explore and demonstrate advanced techniques around the me of data science, APIs, and increasingly LLMs.
As I build stuff here, I write about in my Tip Sheet newsletter. Subscribe here to learn how I build these anchor project components and how you can, too: subscribe to the Tip Sheet newsletter.
This anchor project has some fairly advanced techniques. If you want to build foundational knowledge of FastAPI APIs for AI and Data Science in Python, I wrote a book you should read! Check it out at Hands-on APIs for AI and Data Science: Python Development with FastAPI.
- Building and using APIs for AI and data science uses
- Using Python asynchronous programming techniques to increase performance and reliability in all areas
Here's a big picture view of all of the anchor project components. They show different pieces of an enterprise build around a data source and Python frameworks.
The Air Travel SDK serves as the central integration layer, allowing multiple consumers to reuse the same functionality while minimizing duplicated code.
- Python - pretty much all the code is in Python
- FastAPI - API development platform
- PostgreSQL and Supabase - Cloud PosgreSQL database
- FastMCP - framework for building MCP servers and clients
- Typer - Library to build command line interface (CLI)
- HTTPX - Async library for API calls.
- Scikit-Learn - Python framework for ML model training
- ONNX runtime - Open framework for hosting ML models for inference
- Juypter Notebooks - Every data scientists friend
A command-line interface designed for developers, analysts, and AI coding agents.
The CLI provides a convenient way to search and retrieve flight information directly from the terminal while leveraging the shared SDK underneath.
Repository path:
The central Python package used throughout the project.
The SDK abstracts the underlying API implementation and provides a consistent interface for multiple consumers.
It is used by:
- The Air Travel CLI
- The MCP Server
- Jupyter notebooks
- Streamlit and Gradio applications [FUTURE]
Repository path:
A FastAPI application that exposes flight information through REST endpoints.
The API acts as the primary access layer for flight data and is consumed by the SDK.
Repository path:
A PostgreSQL/Supabase-backed datastore containing processed airline operational data.
The Flights API retrieves flight information from this database layer.
Repository path:
An MCP (Model Context Protocol) server that enables AI assistants and coding agents to interact with the air travel ecosystem through standardized tooling.
Rather than implementing its own database logic, the MCP server reuses the shared SDK.
Repository path:
Jupyter notebooks used for exploratory analysis, experimentation, and prototyping.
These notebooks demonstrate how analysts can work with the same SDK used elsewhere in the project.
Typical activities include:
- Data exploration
- Feature engineering
- Hypothesis testing
- Experimentation
Repository path:
Jupyter notebook demonstrating ML model traing and building an API for inference.
The model trained in this example is pretty naive, so don't look too closely.
But the approach to training and serving inference via API is a solid framework.
Repository path:
Interactive applications built with frameworks such as Streamlit or Gradio.
These applications will provide end-user experiences while relying on the SDK to retrieve data.
Potential use cases include:
- Flight search applications
- Dashboards
- Demonstrations
- AI-assisted experiences
The project uses publicly available airline operational data from the U.S. Department of Transportation's Bureau of Transportation Statistics (BTS).
BTS Data Portal:
https://www.transtats.bts.gov/DL_SelectFields.aspx?gnoyr_VQ=FGK&QO_fu146_anzr=b0-gvzr
Supporting datasets, ingestion assets, and intermediate artifacts used throughout the project.
Experiments involving large language models, prompt engineering, and AI workflows.
Database infrastructure, schema definitions, and supporting scripts.
Clone the repository:
git clone https://github.com/Ryandaydev/anchor_project_air_travel.git
Explore one of the major entry points:
sdk/for reusable client functionalitycli/for command-line workflowsflights-api/for the REST API implementationmcp/for AI agent integrations
The Anchor Project emphasizes several architectural principles:
- One SDK, many consumers -- shared functionality minimizes duplicated logic.
- API-first development -- services communicate through well-defined interfaces.
- AI-ready architecture -- MCP servers and agent workflows are treated as first-class consumers.
- Composable components -- applications can evolve independently while sharing common foundations.
- Educational transparency -- the repository demonstrates practical patterns for modern data, API, and AI engineering projects.