How Machine Learning is Revolutionizing Contamination Detection in Food Factories

The factories that process our food and beverages—surprise, they don’t just come fresh from the farm—must maintain high cleanliness standards to prevent illness. This typically involves traditional methods like petri-dish microbiological monitoring and laboratory testing. A new French startup aims to revolutionize this process using advanced deep-learning algorithms.

Spore.Bio has developed a cutting-edge pathogen-detection system that utilizes optical light to assess surface cleanliness. By comparing light reflection from surfaces that have been in contact with clean and unclean food, the startup can quickly identify contamination.

The innovative approach has recently secured €8 million in pre-seed funding, led by London’s LocalGlobe VC, with participation from EmergingTech Ventures, No Label Ventures, Famille C (Clarins Family Office), Better Angle, Plug & Play Ventures, Entrepreneur First, Kima Ventures, Raise Sherpas, Fair Equity, and Sharpstone Capital.

“Our device captures a sophisticated snapshot of the sample,” explained CEO Amine Raji in an interview. Having previously worked at Nestlé, Raji co-founded the company with CTO Maxime Mistretta and COO Mohamed Tazi, the founder of Gymlib.

Spore.Bio's imaging technology goes beyond what the human eye can discern. “We’ve trained machine learning models to identify the spectral characteristics of bacteria in these images. Our extensive dataset of contaminated and uncontaminated foods and beverages is a major asset, and we've secured contracts with some of the world’s largest manufacturers,” Raji shared.

Currently, Spore.Bio is in its startup phase, with plans to develop a user-friendly device for on-site monitoring. “We’re creating a handheld device that can detect pathogens instantly right on the factory floor, providing near real-time insights into potential bacteria,” Raji stated.

This innovative solution employs a laser that emits light at specific UV-Infrared wavelengths. “Bacteria respond uniquely to this light, and we’ve trained our computer vision and chemometric models to recognize these specific signatures,” he explained. However, Raji was cautious about revealing too much, citing ongoing patenting processes for their advanced photonic technology.

Transitioning from concept to product will involve navigating stringent regulations in food production, particularly the “General Food Law” in Europe. Additionally, any new testing technology must achieve independent certification, which can take 12 to 18 months (ISO 16140 – Microbiology of the food chain).

“We are collaborating with certifying bodies to pursue this certification, though it’s not a prerequisite for initial commercialization,” Raji noted. The product is currently undergoing a rigorous testing and development process to ensure accuracy in pathogen detection.

The market presents significant opportunities for disruption. While factories must regularly test for pathogenic bacteria, current methods have changed little in decades. Presently, samples are sent to offsite labs, resulting in testing delays of 5 to 20 days—slowing decision-making and hindering quick issue resolution. This lag not only affects productivity but also comes with considerable costs.

Spore.Bio aims to provide almost real-time testing results, potentially reducing downtime for food processors. According to Deloitte, the annual cost of downtime in the global food and beverage processing sector is estimated at around $50 billion. (While Deloitte's figure is notable, it’s wise to consider it with some caution.)

Although the product isn’t commercially available yet, Raji mentioned a "waitlist" for their initial prototypes, with plans to launch globally by next year. Their competitors include U.S.-based PathogenDX, which has raised $11.6 million for its various solutions.