Fermentation is one of the trickiest processes to control since it’s all about guiding a living organism. Prior to today’s technology for continuous sampling—and artificial intelligence to make real-time process decisions—making a fine wine needed an expert who could evaluate samples of juice from grapes and from the wine fermentation process, take them to the lab and check the status of key process variables. Based on years of experience, the vintner would make decisions that would, hopefully, make the perfect wine.
Today, thanks to the U.S. FDA—which helped develop the concept of in-line/on-line (aka in-situ) real-time process analytical technology (PAT), and software developers who employed AI and machine learning (AI/ML) to control processes—winemaking is not only an “art” but becoming an exact science—a system that learns what makes a perfect batch of wine and can act accordingly should any variable drift out of limits.
About a year ago, we looked at obtaining better control of fermentation in beer through machine learning and continuous monitoring of process variables in a brew tank, like dissolved oxygen, pH, temperature, etc. This time, we look at a real-time, AI-based fermentation system developed solely for winemaking called Winely.
Winely helps vintners save time, space and costs compared to traditional sampling methods. With this innovation, Co-Founder & CEO Jacob Manning has helped to turn winemaking into a more exact science. We spoke with Manning to learn more about the system, and how it can result in a more sustainable environment.
FE: Where is Winely located and what motivated you to start the company?
Jacob Manning: Winely has research sites in both California and throughout Australia and New Zealand. Our software is currently servicing wine regions in these areas. My co-founder, Abbe Hyde, and I first started Winely about five years ago, out of our shared desire to make a measurable impact through biotechnology.
Five years ago, Abbe and I talked to numerous bio-scientists and were first drawn to the wine industry, after learning it has the most complex fermentation bio-cultures. Quickly, we realized that winemaking is a deeply reactive, and frankly, a dangerous process. We knew there had to be a better way. We discovered the best solution was to capture the right data in real-time; to empower winemakers with choice, visibility and remarkable control.
From there, we developed an AI-powered solution that can extract vast, actionable insights into the fermentation process in real-time, which is an essential task for future alternative protein scaling and development. But we also did the hard work to really dig into the research before launching Winely. We went into this wanting to truly understand fermentation. Now that we have such remarkable insights into the process, we want to help demystify this whole process for those whose businesses center around it.
Winely’s Predictive analysis visualizes how a winemaker’s fermentation will behave over time, plus other winemaking details in real time.
FE: Who makes up your staff?
Manning: Our team is made up of biotechnology experts, software and sensor engineers and machine learning experts.
FE: How are the large commercial wineries today controlling their batches?
Manning: Currently, most commercial wineries still perform manual sampling of their wine batches throughout the fermentation process. This involves sampling a product in the midst of fermentation to gauge whether it has met the winemaker’s standards. The results can still be variable because the inputs—grapes and yeast—change year-on-year. This is why it is so critical for winemakers to keep such a close eye on their ferments. Ensuring it is tracking how they want it to, and they are getting their desired end result.
FE: What’s wrong with how samples and process variable data are taken and recorded?
Manning: Unfortunately, the highly manual method we just outlined above leaves the winemaker as a reactive participant in the process, and candidly, it can be quite dangerous. For example, hydrogen sulfide produced by yeast during fermentation is toxic and people can die from inhaling it. Not to mention the fact that, once a tank of wine has gone bad, a shocking amount of really harsh chemicals are added to it just to make it drinkable.
FE: What are the key process variables in winemaking?
Manning: While there are key attributes that winemakers keep a watch on daily, really what a winemaker is looking for is a holistic understanding of how the ferment is tracking. To get this, there is consideration of biological and chemical attributes. We consider all of these things and how they contribute to overall measures such as fermentation kinetics and yeast vitality.
Winely’s sensors capture data in a tank and measure attributes such as pH, temperature and Brix. In most cases, sensors can be installed by winery personnel.
FE: What defines a perfect batch?
Manning: We’ve seen great wines be produced from the most unlikely data points. This is why real-time monitoring is so critical in this industry, because it gives a winemaker control without needing to follow a prescriptive recipe, which lowers the risk of winemaking but may not yield the most end-quality results. With Winely, they can see what is happening in real-time, and make the decisions that enhance their craft accordingly.
FE: What instrumentation is required for Winely?
Manning: We supply our own instrumentation in the form of Winely’s proprietary, IoT and AI-powered sensors that can be fitted into the tank itself. These sensors have been designed and iterated over the past four years with customers. Our technology system captures data every minute, which equates to millions of data points, and is then shipped to our customers in real-time. The sensors measure attributes such as pH, temperature and Brix, which is sugar and potential alcohol content. As such, we’re measuring and delivering upon areas of the fermentation process, which we call TrueBrix, that the wine industry has never seen before.
FE: How did you develop the software/controls for Winely?
Manning: The software dashboard is cloud based and wineries can access it anywhere they have an internet connection. Wake up in the morning and check it on your phone—whatever you fancy! We give winemakers total control to set the alerts and parameters based on what they care about and how they want their ferments to behave. Many do this by grape varietal or even yeast type.
FE: How have you embedded AI/ML into Winely?
Manning: We’ve trained the fermentation algorithms with millions of data points and years of research to enhance the winemaker’s control of their fermentation process. Using these algorithms, we’ve developed Winely’s “Predictive Analysis” and “Scenario Planning” features. Predictive Analysis visualizes how a winemaker’s fermentation will behave over time, in the future—which will help them track if their fermentation is on schedule. Scenario Planning is an excellent companion tool that can prevent costly mistakes, as it allows the winemaker to add hypothetical actions to their batch and see the results that it will produce.
FE: Have any wineries been using Winely?
Manning: Absolutely. Many of the most advanced wineries worldwide have been benefiting from our solution, such as Maori Point Wines in New Zealand. When a winery uses our technology even once, they understand it. Their eyes light up, and they ask if we can put one in every single tank. This is because they know that the ability to sense one problem in one tank with one alert can enable them to react 12 hours earlier and offers enormous savings potential as a consequence. However, when you also factor in the safety protections, the peace of mind, the less frantic atmosphere, the real-time choice and the richer vintages, this technology can produce for deeper control, innovation and creativity. The values and applications are almost endless.