Growing Ideas: Precision milk sensors lead the whey to the future

Shedding light on the health of the herd

Fresh from picking up the Judges’ Choice award at FoodBytes! in New York, SomaDetect’s CEO Bethany Deshpande explains how what began as a father-daughter project could end up revolutionizing milk quality and herd health for the dairy industry.

Dairy accounts for approximately 10 percent of the world’s protein and is a critical part of many people’s everyday diet. It’s also resource hungry, which is why it’s never been more important to improve the efficiency and sustainability of the industry.

Delivering those improvements and ultimately making better milk is the aim of SomaDetect, a New Brunswick, Canada-based start-up that is bringing the Internet of things and artificial intelligence to the humble dairy farm.

By combining data, deep learning and light scattering detection technology, SomaDetect can monitor the health of the herd and the quality of milk it produces more accurately and time critically than ever before.

Lights, data, action!

SomaDetect CEO Bethany Deshpande explains how the idea came about: “My dad is a biophysicist who has been working on applications of light scattering for my entire life. I grew up watching his inventive and entrepreneurial spirit at work. He started using milk to calibrate his sensor and realized he could use the technology to detect all kinds of things in milk and other fluids.

“That’s when I became involved. It started out a way of helping him get this technology out into the world, but I found a compelling problem to solve and a huge market. When we added machine learning and artificial intelligence, we found we could measure even more compounds than originally envisaged.”

SomaDetect sensors installed at a Cornell University milking facility.

The SomaDetect sensor fits into standard dairy milking equipment. It works by directing a beam of light through the milk, which hits particles that cause the beam to change direction and scatter. This light is absorbed or reflected, changing the intensity of the beam according to the kind of particle it hits.

This data is then fed through deep-learning algorithms that predict the presence and concentrations of major compounds in the raw milk. From this, SomaDetect is able to determine fat and protein content, and trace antibiotics and progesterone, making it possible to ensure milk quality and diagnose several major dairy diseases, including mastitis, ketosis and acidosis – in real time.

“Farmers get measurements from every cow at every milking in real time”

- Bethany Deshpande, SomaDetect

Rapid reaction

The game-changer for the industry is that this technology allows dairy farmers to act on the data they receive quickly enough for it to have a beneficial effect on herd health and milk quality.

“We typically refer to it as addressing a cold rather than pneumonia,” says Deshpande. “Right now, dairy farmers get this data through conventional systems every six weeks. With SomaDetect, they get measurements from every cow at every milking in real time, delivered to whichever device they choose.

“What’s more, since no laboratories are involved, there is no cost per measurement and no need for using reagents or chemicals either.”

The benefits for farmers are manifold. Not only can they identify sick cows and take them out of the milking line, they can monitor new feeding regimens, eliminate antimicrobial contamination and share data with vets. In short, SomaDetect allows dairy farmers to optimize every aspect of their operations.

(Machine) learning curve

According to SomaDetect, dairy farmers will see a substantial return on investment within the first year of using the system. But selling the benefits of deep learning to dairy farmers has been a challenge, Deshpande recalls.

“Most folks who understand dairy don’t necessarily know about deep learning and vice versa, so it’s been a fun challenge bringing the two worlds together on the farm,” she says. “I had a lot to learn about dairy farming when we first started out. There’s so much complexity around running a profitable operation, from raising cows and caring for them when they are sick to milking them and producing food – and I still have so much more to learn.

“Most dairy folks don’t know about deep learning and vice versa”

- Bethany Deshpande, SomaDetect

“But we have been welcomed enthusiastically by our early adopters, who are highly motivated to make their farms more sustainable. Collaborating with them has meant we’ve managed to really hone down our story to its most critical elements.”

It’s an exciting time for SomaDetect and its thirty early users. The first devices are currently going out to farms.

“We’ll be working closely with our early adopters over the next several months,” adds Deshpande. “That involves finalizing algorithms, measuring the ROI for farmers and ensuring what we deliver for them is of the maximum value possible.”

FoodBytes! is a Rabobank initiative connecting agrifood start-ups with industry leaders and investors. Applications are open now for FoodBytes! San Francisco on March 27, 2019! Learn more.