Cisco data scientists work with nonprofit partner Replate to improve food recovery and delivery to communities in need – Cisco Blogs
The Transformational Tech series highlights Cisco’s nonprofit grant recipient that uses technology to help transform the lives of individuals and communities.
Artificial Intelligence (AI) and Machine Learning (ML) are utilized in many different industries. AI and ML create more efficient virtual healthcare visits and more intuitive online education platforms. They enhance agriculture through IoT devices to monitor soil health, and devise new ways for people to access banking and other financial services.
This type of technology can also be used to improve services that nonprofits provide to local communities. At Cisco, we have a proven track record of supporting nonprofits through our strategic social impact grants along with a strong culture of giving back. Cisco’s AI for Good program brings these values together by connecting Cisco data science talent to nonprofits that do not have the resources to use AI/ML to meet their goals.
Cisco AI product manager and former data scientist Arya Taylor leads the AI for Good program. Arya shared, “AI for Good is specifically dedicated to the data science community at Cisco. We heard that a lot of data scientists want to apply their skills to a problem for good.”
The AI for Good team constantly works to grow its network of nonprofit partners by engaging with the team that manages Cisco’s social impact grants and by reaching out to nonprofits directly. One of the organizations that AI for Good volunteers support is Cisco nonprofit partner Replate. Based out of Oakland, California, Replate reduces food waste through a digital platform that makes it easy for companies to schedule on-demand pickups for their surplus food. Replate’s food rescuers bring donated food to nonprofit partners who distribute it to people of all ages and backgrounds who are experiencing food insecurity.
Cisco data scientists use ML to forecast food supply and optimize Replate’s operations
Cisco’s AI for Good team spent six months working with Replate to develop a model that can forecast food supply to maximize food recovery and optimize their operations. Replate’s staff met with Cisco’s AI for Good team via WebEx to share more about their method of food recovery. Cisco data scientists first assessed the scope of Replate’s needs and learned how they could best apply their skills in ML to make an impact.
This Cisco AI for Good project was led by data scientist Aarthi Janakiraman, who also served as cause champion–which means she led the project from start to finish to ensure the project’s success. Other members of the project included data scientist Idris Kuti and ML operations expert David Meyer. The team looked at how Cisco’s machine learning models would allow Replate to predict surplus food supply within their donor network.
Because Replate offers a variety of donor plans to their partners, it can be challenging to calculate availability and capacity. As a result, Cisco’s data scientists developed an ML model that could predict the total pounds of food each donor would contribute on any given day. This more accurate prediction helps Replate’s food rescuers, who deliver the food, as well as the nonprofit organizations that rely on meal delivery.
“Before our project started,” Arya explained, “Replate was using a rules-based model with different thresholds that would determine the estimated amount. But there’s no single threshold in machine learning that you can apply to every single donor; it just becomes more personalized to that donor and evolves as more data is collected. So, it works more like our brain, rather than a static generalization.”
Aarthi gave an example: “Let’s say there is a donor for Replate, and they tell us that next Friday they will be able to provide 60 trays of food. This number is often a skewed estimate–donations are typically from grocery stores, corporate cafeterias, or farmer’s markets, who may not be able to provide an exact prediction due to the variability of consumption. Our model will take in different information about the donor and estimate a more accurate donation amount in pounds. That estimation will go into Replate’s algorithm and match the food rescue task to the correct driver.”
By incorporating machine learning models, Replate can also predict donation volume for existing and new partners. The volume of a new donor’s first pickup will be predicted based on data from donors in similar regions or industries. “Such forecasting will make a significant difference in our operations and allow us to better fulfill our mission,” said Mehran Navabi, senior data scientist at Replate. “Replate will implement these models into our codebase and integrate them within our existing routing algorithm. The algorithms will coalesce to automate driver-dispatches for each donor’s pickup.”
Cisco and Replate: Working together to create lasting change
Replate’s team met with Cisco data scientists for biweekly progress reports throughout the project lifecycle and discussed how they could advance their platform’s technological capacities. The models that the AI for Good team created will enable smarter dispatching, which will allow a greater volume of food to be recovered and delivered to communities in need.
According to Mehran, one challenge for Replate is meeting the different needs and expectations of their nonprofit partners who serve diverse populations with varying capacities for food storage and meal distribution. Having a model to forecast food supply can reduce waste and help Replate connect food delivery tasks to the correct drivers to ensure as much food as possible will be given to those in need. The project may even increase the amount of surplus food that can be recovered by giving Replate the information needed to make smarter, predictive dispatching decisions.
Now, Cisco’s AI for Good team is handing over the project to Replate and will leave them with a maintenance plan which will allow them to retrain the model on Google Cloud Platform. They also built out a service that will track the model’s accuracy, so any adjustments can be made as time goes on.
“Working with Cisco’s AI for Good team was incredible,” said Mehran. “Their team was professional and knowledgeable. And overall, their communication was excellent. The partnership enabled Replate to build a fruitful and beneficial connection with the Cisco team and foster new approaches to the way we collect and interpret data.”
Share:
