#  Project 25009 AI-Driven School Bus Stop Safety Evaluation 

 



**Project Title**

AI-Driven School Bus Stop Safety Evaluation

**Zifu Wang**, Postdoc Research Fellow, Harvarad CGA  
[zifu\_wang@fas.harvard.edu](mailto:zifu_wang@fas.harvard.edu)

  
**Project Description**

This internship will support the development of an intelligent evaluation system for school bus stop safety using geospatial analysis and artificial intelligence. The intern will integrate traditional spatial verification methods with computer vision, large language models (LLMs), and deep learning to identify and assess potential hazards. The goal is to automate and scale the assessment of bus stop safety using multimodal data (text, imagery, maps).

  
**Tasks and Responsibilities**

1\. Data Acquisition & Preprocessing

1\) Collect bus stop location data and contextual metadata (e.g., traffic features, environmental risks).

2\) Extract visual context using APIs such as Google Street View or satellite imagery services.

2\. AI & Deep Learning Tasks

1\) Use computer vision (CV) models to detect road signs, curbs, sidewalks, obstructions, and pedestrian zones from street-level images.

2\) Apply deep learning models (e.g., CNNs, YOLO, Segment Anything) for object detection in urban environments.

3\. LLM Tasks

1\) Use large language models (e.g., GPT-4, LLaVA) to interpret inspection forms, parse natural-language safety descriptions, and generate hazard explanations.

2\) Build prompt-based or RAG systems to support automated reasoning about bus stop compliance.

4\. Integration and Automation

1\) Develop pipelines to combine CV, LLM outputs, and rule-based evaluations.

2\) Visualize safety scores and detected hazards using QGIS or web-based mapping tools.

  
**Minimum Qualifications**

1\. High school or undergraduate student with strong experience in at least two of the following areas:

1\) Python programming and data analysis (e.g., pandas, Jupyter)

2\) Computer vision or deep learning frameworks (e.g., YOLO, OpenCV, PyTorch)

3\) Use of large language models (e.g., OpenAI GPT, Hugging Face Transformers)

4\) GIS tools or spatial data platforms (e.g., QGIS, OpenStreetMap, Google Earth Engine)

2\. Demonstrated interest in AI applications for urban safety, transportation, or civic technology

3\. Ability to work independently and follow technical documentation

**Terms of the Project**  
 one year  
  
**Key Words**

bus routing, stop evaluation