The competition invites participants to develop machine learning (ML) and deep learning (DL) models that forecast and detect food safety risks across the supply chain using curated datasets from both academic and industry sources.

Leveraging the Cornell Food Safety ML Repository, newly contributed datasets from Agroknow’s Foodakai platform, and publicly available computer vision datasets generously permitted by Dr. Yi from Michigan State University, competitors will apply AI/ML tools to real-world data to address contamination risks in food production environment, food risk in global food trade, and rapid detection of bacteria.

This initiative aims to foster collaboration across domains, drive innovation, and demonstrate the potential of predictive analytics to protect public health. This competition is supported by the AI Institute for Next-Generation Food System (AIFS), Food4AI from University of Guelph, and IAFP Student PDG.

Winners may have the opportunity to attend the Annual Meeting of the International Association for Food Protection (IAFP).

Competition Objectives and Tasks

Main Objectives:

• Build predictive models for identifying or forecasting food safety hazards in various environments and supply chain stages.
• Generate actionable insights for food safety surveillance and risk mitigation.
• Define the benchmarks for the performance of ML/DL models in food safety

Tasks:
 Participants will:

• Select one or more datasets from the Cornell Food Safety ML Repository, Agroknow’s Foodakai risk intelligence platform, and computer vision-based pathogen detection datasets.
• Develop and document an end-to-end ML/DL pipeline that includes:
  • Data preprocessing and feature engineering
  • Model training and validation
  • Interpretation of model outputs
• Address one of the following challenge themes:
  • Predicting pathogen presence in food production environment
  • Forecasting food safety incidents based on Agroknow’s historical global food safety incidents (recalls, border rejections) and publicly available supply chain data (price data, trade data, laboratory test results, weather data, etc)
  • Image-based bacterial classification and detection

Judging Criteria

A 1–2 page written report is due on March 1, 2026, and will be used to select the finalists for the online judging round, which will determine the top three winning teams. The report should include a concise description of the problem addressed, methods used, key findings or insights, and must include the code or repository link for the developed pipeline or model to ensure reproducibility. Final submissions will be evaluated through online presentations.

The submissions (report and online presentation) will be evaluated based on the following rubrics:

• Performance Metrics:
  • Classification tasks: AUC, Prediction-Recall Curve, F1-score, Sensitivity, Specificity
  • Forecasting tasks: MAE, MSE, RMSE, SMAPE, or other suitable timeseries forecasting metrics
  • Object detection tasks: mAP, IoU, Precision, Recall
• Integration and Innovation:
  • Effective use of multiple datasets
  • Novel approaches to feature selection or model architecture
• Interpretability and Actionability:
  • Clarity of insights and potential use in real food safety monitoring
• Code Quality and Reproducibility:
  • Well-documented, modular, and reusable code shared through open-access platform
• System Requirements:
  • GPUs, CPUs, RAM and ROM specifications or other system requirements for reproducibility

Datasets to be used

Participants can choose from:

GIS-based pathogen presence prediction:

• Listeria in soil
• Salmonella and Campylobacter in chicken
• Fecal contamination in surface water

Agroknow / Foodakai Datasets:

• Historical food Incidents for specific products

Computer Vision:

• Microscopy for bacterial classification under varying conditions
• Biosensing for pathogen classification and localization
• Hyperspectral microscopy for Salmonella serovar classification