HiLT: A Library for Generating Human-in-the-Loop Data Transformation GUIs
summary
HiLT is a Python library that generates web GUIs, backend code, and a database from a pipeline of Stage abstractions (each stage is a Python function that defines the UI and the data transformation). Developers build a staged pipeline for tasks like the wildfire→county→join flow, producing a working tool in under an hour. In a user study HiLT outperforms Streamlit: participants complete more tasks overall and are 2.8× more successful on data-transformation subtasks, showing the stage-first approach reduces state and schema errors.
AI-written summary · about
chapters (8)
- 0:00Introduction to HiLT
- 0:44Human-in-the-loop data tasks
- 2:48HiLT capabilities and examples
- 3:50Challenges in GUI building
- 5:17HiLT’s stage abstraction
- 6:21User study results
- 7:45Lessons for GUI frameworks
- 8:07Q&A session
related talks
01h 58mHow to create beautiful interactive GUIs and web apps
16mDemocratizing Data Pipelines: Introducing the AI-Powered ETL Assistant
04mAbusing text/template for data transformation
37mAdrian Brudaru | Data Track | Compass AI & Data Summit 2025
39mHow to build stunning Data Science Web applications in Python with Taipy
38mBuilding Data Pipelines in Python
26mBuilding Data and Machine Learning Applications using Gradio
25mHumans helping machines help humans run machines