Founded by Karan Kumar Singh, Nikita Gajbhiye and Abhileen Pandey, the initiative is combining computer vision, geospatial technology and community participation to help surplus food reach people more efficiently.
New Delhi: As artificial intelligence becomes increasingly associated with commercial automation, a team of young Indian technologists is applying it to a more fundamental challenge: connecting surplus food with people who need it before the food becomes unsuitable for consumption.
Founded in 2023 by Karan Kumar Singh, Nikita Gajbhiye and Abhileen Pandey, ZeroHungerAI has developed a technology-driven food redistribution platform designed to reduce food waste and improve access to available meals. The initiative has contributed to the distribution of more than 2,000 meals in Delhi and works with a wider team across technology, operations and community outreach.
The platform initially focused on location-based food discovery. Donors such as restaurants, organisations and individuals could list surplus meals by providing information about the food type, quantity, preparation time, condition and pickup location. People searching for food could then discover nearby listings within a defined radius without creating a recipient account.
ZeroHungerAI has now expanded this platform with AI-assisted capabilities intended to improve listing quality, prioritise time-sensitive donations and support more efficient collection.
AI-Assisted Food Image Analysis
ZeroHungerAI has deployed an AI-assisted computer-vision pipeline that analyses food photographs uploaded by donors. The system suggests possible food categories, compares visible condition indicators with donor-reported information and flags uncertain cases for human review.
The model is evaluated through a reproducible testing process and served using ONNX inference, allowing it to provide confidence-based suggestions efficiently. It does not publish conclusions independently. Donors must review and confirm the suggested information before their listing becomes available on the platform.
This human-in-the-loop design is intended to reduce incomplete or inconsistent listings while ensuring that people remain responsible for the final information. The system does not present image analysis as a certification of freshness, hygiene or food safety, since these cannot be reliably determined from a photograph alone.
The platform also limits unnecessary storage and processing of uploaded information to support a more privacy-conscious approach.
Prioritising Urgent Donations
Food redistribution is highly time-sensitive. A donation available now may no longer be suitable several hours later, making prioritisation an important part of the collection process.
To address this challenge, ZeroHungerAI has introduced a versioned rescue-priority engine that considers factors such as the time remaining before a listing expires, food quantity, reported condition, distance and collection feasibility.
Rather than treating every listing equally, the system helps identify donations requiring faster attention. Its versioned design allows the team to evaluate and improve the prioritisation logic over time while maintaining a record of how individual recommendations were generated.
The priority engine works alongside a pickup-route planning system that helps rescue partners coordinate collections more efficiently. By considering the locations and urgency of multiple donations, the system can suggest practical pickup sequences and reduce avoidable travel.
These recommendations serve as decision-support tools. Volunteers, NGOs and rescue partners can review them alongside real-world considerations such as transportation availability, local accessibility and storage capacity.
Connecting Technology with Community Outreach
ZeroHungerAI has also signed a Memorandum of Understanding with Delhi-based Hamari Pahchan NGO. The collaboration brings together Hamari Pahchan’s community outreach experience and ZeroHungerAI’s technology-driven platform to connect surplus food with communities more effectively.
The partnership reflects an important aspect of the initiative’s approach: artificial intelligence alone cannot redistribute food. Successful collection still depends on donors, volunteers, NGOs and community networks working together.
Technology can, however, make that coordination more informed. Location-aware discovery can help identify nearby donations, computer vision can assist donors in completing listings, the priority engine can highlight urgent cases, and route planning can support faster collections.
Developing a Responsible AI System
For the ZeroHungerAI team, the objective is not to automate sensitive decisions without oversight. Its AI capabilities are designed to provide suggestions, identify uncertainty and assist human decision-making.
This distinction is particularly important in food redistribution, where an incorrect assessment may have consequences for recipients. The system therefore avoids claiming that a model can certify food safety based on an image. Donor confirmation and human review remain essential parts of the process.
As the platform collects more operational information, the team also plans to study patterns in food availability, demand, collection times and unsuccessful pickups. These insights could eventually support demand forecasting, surplus prediction and geospatial hotspot detection across urban communities.
With more than 2,000 meals distributed, an institutional partnership supporting its outreach and new AI-assisted capabilities deployed within the platform, ZeroHungerAI is moving beyond basic food discovery toward a more intelligent redistribution system.
The initiative demonstrates how artificial intelligence can be used not simply to automate processes, but to improve coordination around a resource that already exists and must reach the right place at the right time.
Website: www.zerohungerai.com
