1 of 7

A Data Fabric For Social Good?

Amarnath Gupta

University of California San Diego

2 of 7

What This Talk is Not About

National Science Data Fabric Meeting, 2023

How do we Democratize the Data Fabric for People?

3 of 7

Is this a Science Data Fabric Problem?

  • A community-conscious philanthropist passionate about California’s food insecurity problem asks
    • Where are the food desert areas near me?
    • How do I identify and contribute toward two qualified local entrepreneurs interested in providing fresh food to their communities?
    • What level of funding would support these entrepreneurs?

National Science Data Fabric Meeting, 2023

3

4/11/2023

NSF Convergence Accelerator Track Problem

4 of 7

Where is the Nearest Food Desert?

  • A census tract/block-group collection that meets both low-income and low-access criteria including:
    • The poverty rate is greater than or equal to 20 percent OR median family income does not exceed 80 percent statewide (rural/urban) or metro-area (urban) median family income;
    • At least 500 people or 33 percent of the population located more than 1 mile (urban) or 10 miles (rural) from the nearest supermarket or large grocery store.
  • A food desert that has an over-abundance of addictive, unhealthy food is called a food swamp
  • Where does the computation happen? Where are the results stored?

​

National Science Data Fabric Meeting, 2023

4

4/12/2023

“Food deserts are geo­graph­ic areas where res­i­dents have few to no con­ve­nient options for secur­ing afford­able and healthy food”

5 of 7

Find Qualifying Entrepreneurs

  • Recommendation Problem
    • Funder’s interest profile and entrepreneur’s need profile
  • Qualifying process
    • Veracity of claims aided by LLM
    • Profile similarity using LLM
    • Chances of success – metrics based on demand and competitiveness
  • Domain-specific Fine-tuning and Request-specific Prompt Engineering?
  • Integration with external data?
  • Conversational Response Time?

National Science Data Fabric Meeting, 2023

5

4/12/2023

A loan of $11,000 helps my Argentinian and Uruguayan food business buy inventory, new equipment and pay for employees.

​

Arlenne's story

My name is Arlenne. I was born in Oaxaca, Mexico, and in 2001 I moved to Milwaukee, Wisconsin, with my parents. … In 2011 we moved to Milwaukee to be close to my family. … We decided to stay but with the goal of opening our own Argentinean food restaurant; … this cuisine is something that did not exist in Milwaukee. … Hence, we started the restaurant El Gaucho Grill, a traditional Argentinean food restaurant. …

Business description

Our business opened its doors on June 14, 2022, with a lot of effort and a great economic investment. The Gaucho Grill specializes in Argentine and Uruguayan food where we sell classic Argentine empanadas and Argentine asado. We elaborate our dishes all at home from the meat that is cooked as in Argentina to our desserts alfajores that are made at home with Argentine dulce de leche. …

What is the purpose of this loan?

The reason I am applying for this loan is to maintain and expand our business, help me in the purchase of equipment and inventory, as well as marketing and the rest of the community outside of our city to spread the word about us. I want to keep fighting for this because this is my dream.

6 of 7

How Much Funding?

  • What’s the average/min/max startup / operating cost (for a period of “N” months) of an entrepreneur with need-profile “P” in region “R” at this time?
  • On-demand data access and time-to-life caching of external data sources
    • Where is the provision for on-demand data access at scale?
    • Where and how is the cache managed?
  • Combination of OLAP-style analytics and ensemble of cost-prediction models
    • Where is the model store? How does one perform model selection?
    • Can this be performed on edge-devices given a minimal hardware configuration?

National Science Data Fabric Meeting, 2023

6

4/12/2023

7 of 7

The Final Questions

  • Democratization of data fabrics for social good requires a ground-level needs assessment process from user-classes in the society
    • How is this needs assessment baked into the data fabric design?
  • Social good platforms should not have to build the entire architecture
    • How are the application interfaces and workload SLAs designed?

​

National Science Data Fabric Meeting, 2023

7

4/12/2023