Meeting Notes - SarahE-Dev/bloom-housing GitHub Wiki

Meeting Notes – 4/22/2025

Participants: Bryan S., Sarah E., Ryan P., Darien D., and Nashid M.

Agenda

  • Data Anonymization
  • Features -> Race
  • Utilizing Docker

Discussion Summary

  • Data Anonymization: Question: Where should data anonymization occurβ€”prior to the API call to the microservice, or within the microservice itself? Sarah E. and Darien D. raised concerns about error handling and testing the form submission before the API call, suggesting that these processes be addressed early in the data flow.

  • Feature Inclusion – Race: Question: Should Race be included as a feature in the model? In efforts to mitigate bias, how might the inclusion of Race impact the "at-risk" score/result? Ryan P. and Darien D. expressed opposition to using Race as a model feature. Sarah E., Bryan S., and Nashid M. recommended documenting the impact of Race within the two datasets on the model’s outcome in order to assess its influence more objectively.

  • Utilizing Docker Question: Should we use Docker to containerize the Microservice? Bryan S. raised the question of confirming Docker usage for the microservice. Sarah E. had already incorporated Docker in the model mockup. Darien D., Ryan P., and Nashid M. agreed that using a Docker container would facilitate easier integration.

Decisions Made

  • βœ… Decision 1 - The group agreed to anonymize data prior to the API call to the microservice.
  • βœ… Decision 2 - Sarah E. and Darien D. will address error handling and testing of the form submission at a later date.
  • βœ… Decision 3 - Bryan S. and Nashid M. will compare datasets to document the potential impact of including Race as a model feature.
  • βœ… Decision 4 - The group agreed to use Docker to containerize the microservice for easier integration.

Action Items

  • --- Error Handling & Testing – Sarah E. and Darien D. – Due Date: TBD
  • Race Feature Documation & Determination – Bryan S. and Nashid M. – Due Date May 1st

Meeting Notes – 4/27/2025

Participants: Bryan S., Sarah E., Ryan P., Darien D., and Nashid M.

Agenda

  • Form Submission
  • Slide Deck
  • MVP
  • Visualizations
  • Model

Discussion Summary

  • Form Submission: Question: What is the reasoning behind requesting additional demographic questions (sex/education level/LGBTQ status)? Bryan S. presented the "Shaping Form Submission" Jupyter Notebook analysis, which provided insights into the potential impact of asking these additional demographic questions. Following the presentation, Sarah E. inquired about how the findings were framed and presented.

  • Slide Deck: Question: N/A Darien D. walked everyone through updates to the slide order and content. Initially, Sarah E. and Darien D. wanted to retain the slide titled "The Disproportionate Impact on Vulnerable Communities." However, Nashid M. questioned this decision, arguing that if the team had decided not to use race/ethnicity as a feature in their analysis, it seemed inconsistent to include this information in the presentation.

  • MVP: Question: What information should we display to the Applicant? A discussion emerged regarding how to present risk assessment results to applicants: Sarah E. advocated for showing the "at-risk" score or Boolean value directly to applicants, expressing concerns about oversimplifying the model's results when communicating with other stakeholders. Nashid M. countered that providing a raw score or Boolean value to applicants would offer little practical value. Bryan S. and Darien D. suggested a middle-ground approach for an effective MVP: returning two different sets of resources based on the "at-risk" score/Boolean value without necessarily showing the score itself. Sarah E. raised additional concerns about aggregating resource data and questioned whether presenting the model's results should be treated as a proof of concept. Nashid M. proposed a compromise for the MVP presentation: demonstrating both approaches to show how the model works while explaining the reasoning behind displaying specific sets of resources to different applicants.

  • Visualizations: Question: Should a separate container be created to house the data for visualization? Bryan S. initiated a discussion on potential visualization approaches for different stakeholders. Sarah E. proposed housing anonymized data to support these visualizations. In response, Nashid M. raised concerns about the storage requirements for maintaining both application data and anonymized data, as well as the additional computational resources needed to run aggregations on the anonymized dataset after initial model results.

  • Model: Question: N/A After fitting the dataset to multiple different models, Bryan S. indicated he is leaning toward implementing a classification model for the initial framework, while remaining open to further testing and refinement.

Decisions Made

  • βœ… Decision 1 - The group agreed to table the request for adding additional questions to the application form.
  • βœ… Decision 2 - The group agreed to present resources based on the "at-risk" results and the actual results via console log during the MVP presentation.
  • βœ… Decision 3 - The group will iterate on the slide deck to accommodate some of the new ideas and concepts.
  • βœ… Decision 4 - The group agreed to table the visualization concept until after the MVP presentation.

Action Items

  • Slide Deck (WIP) - Group - Due Date May 4th

Meeting Notes – 4/29/2025

Participants: Bryan S., Sarah E., Darien D., and Nashid M.

Agenda

  • Feature Justification Documentation
  • Presentation Flow

Discussion Summary

  • Feature Justification Documentation: Question: Where should the feature justification documentation be stored? Nashid M. proposed storing it in the Wiki, while Sarah E. recommended placing it in the documentation folder within the repository.

  • Presentation Flow: Question: N/A The group agreed on the need to refine the presentation flow. Sarah E. asked when the MVP should be presented, and Bryan S. inquired about who should present it. Darien D. suggested using a story to hook the audience, and Bryan S. proposed framing the model as a tool to capture missed opportunities.

Decisions Made

  • βœ… Decision 1 - The group agreed that Feature Justification Documentation should be pushed into the repository.
  • βœ… Decision 2 - The group agreed that Sarah E. will present the MVP.
  • βœ… Decision 3 - The group will continue to iterate on the slide deck to accommodate some of the new ideas and concepts.

Action Items

N/A

Meeting Notes – 5/27/2025

Participants: Bryan S., Sarah E., Ryan P., Darien D., and Nashid M.

Agenda

  • Admin Access (Jurisdiction)
  • Google Analytics 4
  • Clustering Modeling

Discussion Summary

  • Admin Access (Jurisdictions): Question: In building out admin access, how should the jurisdictional access hierarchy be framed out? Sarah E. posed this question to Colin B. and Adrien Y. (Exygy mentors). Colin B. provide a demonstration of how the hierarchy is setup within Bloom Housing application. Colin B. suggested that Admin access should start at the Administrator Level and work its way down. He provided an example of aggregating all applications then breaking them out to those who are applying within current jurisdiction to those who are applying outside of their jurisdiction.

  • Google Analytics 4: Question: How is Google Analytics 4 (GA4) used within Exygy products, and could Team 4 have access to Exygy's GA4 account? Nashid M. referenced a previous conversation that the team had with the Exygy mentors, where they spoke to using GA4 for analytics and presentation purposes. Darien D. ask if we could have access to Exygy's GA4 account to help build out a feature of tracking most used resource links.

  • Cluster Modeling: Question: Bryan S. share clustering modeling documents with the team.

Decisions Made

  • βœ… Decision 1 - Darien D. will create a free GA4 account to track resource usage.
  • βœ… Decision 2 - As a team we decided to use the Clustering Modeling to generate the "at-risk" score.

Action Items

N/A