Energy Efficient Household Rewiring
Appliance Lookup
Project Description
In the next few decades, we’ll replace 1 billion household machines and appliances, and it's critical for the climate that we replace them with clean electric versions instead of the fossil fuel-based appliances of the past. To help consumers choose and compare these machines, take advantage of tax incentives, and assist contractors in finding the right fit, we need a centralized platform.
During the Spring 2024 semester, our team of four is diligently working on this project as part of the Computing for Good course at Georgia Tech. We are collaborating closely with Rewiring America to develop this essential tool.
Project Goals
Create a database and API for household electric appliances such as furnaces, heat pumps, and water heaters, and use it to build a consumer-facing web app.
Tasks:
- Collect nomenclatures: the organizational structure of appliance model numbers and what they represent.
- Determine how to represent it in a database.
- Create a parser that searches the database for matches.
- Construct a front-end that takes applicance model numbers, finds matches, and unpacks it for a user.
Team Members
Dan Turcza
Backend Development
Along with Kevin, Dan will focus on the data discovery workstream, which focuses on where appliance data sources are located. The goal is to have a comprehensive understanding of all makes/models of target appliances, validated by other aggregators in the space. Along with Chris, Dan will also focus on the data processing workstream, which is concerned with turning those data sources into usable data that fits our schemas. This will involve ingesting web resources and processing them to extract the data we care about. Dan is also the team liaison with Rewiring America since he has a preexisting relationship with them.
Kevin Ferri
Backend Development
Along with Dan, Kevin will focus on the data discovery workstream, which focuses on where appliance data sources are located. The goal is to have a comprehensive understanding of all makes/models of target appliances, validated by other aggregators in the space. Along with Ricky, Kevin will work on frontend dev and the UX to build a compelling and accessible web application that users can use to explore the data that we've collected. A stretch goal for this workstream is integration with other datasources such as financial incentives in order to allow for more complex applications.
Ricky Cheah
Front-End Development
Along with Kevin, Ricky will work on frontend dev and the UX to build a compelling and accessible web application that users can use to explore the data that we've collected. A stretch goal for this workstream is integration with other datasources such as financial incentives in order to allow for more complex applications. Ricky is also the main author of this website.
Chris Yang
Backend Development
Along with Dan, Chris will focus on the data processing workstream, which is concerned with turning those data sources into usable data that fits our schemas. This will involve ingesting web resources and processing them to extract the data we care about. Chris will also be the primary database developer, concerned with creating a viable database and serving API so that clients can make use of the data that we've collected.
Team Weekly Updates
Dan
Highlights: Added a new stage to incorporate PDF text, eval framework, documentation for P9, got our water heaters into the database.
Challenges: LLM struggles with some fields like dimensions, but it is surprisingly quick to correct. The interface in and out of csv is still a bit awkward because of the nested fields (like dimensions) and the fact that we lose type information going to CSV, so then the numbers appear as strings in the JSON which does not match our schema
Next Steps: Add more appliances into the database (heat pumps next.)
Kevin
Highlights: Made changes to csv-to-json script, tried playing around with the new pipeline but mostly ended up overfitting, small changes to the readme.
Challenges: Overfitting when prompt engineering.
Next Steps: Complete background for P10 and add slides to final presentation.
Ricky
Highlights: Finished adding hvac, and cleaned up the code for the frontend. Removed unused functions and files, named the files more consistently, and factored out helper functions into a helper file. Added part of P9 about frontend.
Challenges: Implementing tests.
Next Steps: Add relevant frontend details to P10. Finish adding frontend tests.
Chris
Highlights: Backend developer documentation on GitHub and P9. Added some tests on the backend, error handling implementation in progress.
Next Steps: Wrapping up the error handling implementation, cleanup the code and add more descriptive documentation ready to handover to Rewiring America. Testing with the data generated from the pipeline. Final presentation and P10