Machine Learning Series: Decontaminating Beer
Beer is a $111.4 billion market in the US alone. The brewing and fermentation process at a large scale is very complicated and has not changed for hundreds of years. While the tech revolution has greatly influenced manufacturing and production globally, it has left the beer industry undisturbed. Large breweries have always been capital intensive and reliant on manual labor. A group of Cal students are hoping to use an IoT device to improve the brewing and fermentation process.
Fiat Lux Labs was founded by four Cal students and was part of Free Ventures’ Batch IX in Fall 2017. They built a sensor technology platform that tracks metabolic flux, contamination, and bio mass in the biochemical processing industry. Examples of industry applications are brewing, pharmaceutical manufacturing, and synthetic biology. Using their platform, companies can increase profit margins by 10–15%.
Part 1: The Idea 💡
Q: How did you come up with the idea to decontaminate beer?
Yash: “We were all part of a biochem lab at Cal and had toyed with the idea of making our own beer for fun. I would set up fermentation batches and manually take samples each hour for 48 hour cycles to test for contamination. We quickly realized that contamination was a big problem and began brainstorming solutions. After considering many ideas, we decided building a sensor for a fermentor was the best way to approach the problem. It was a long process, but that’s how we shaped the science into an actual product.”
Q: When did you think about turning the idea into a company?
Yash: “My friend who worked at Anheuser-Busch invited me to a startup conference they sponsor. I started talking to their director of innovation and he became really interested in our personal project. I took that as a sign that the idea had real monetary value, which is when I started looking into how to start a company.”
Q: What kept you motivated throughout the process of building the company?
Yash: “A couple of things to keep in mind is that you always want to have a vision of the end goal for the company. For us, we envisioned that although we’re starting in brewing, we wanted our tech to be in almost every industry involving biochemical processing. We also didn’t get any investment the first 6–8 months. It’s important to talk to a lot of people before you get that fist paycheck and piece of validation. The last thing is to be flexible. We started with a completely different idea. But if you stay really flexible to possible new paths and you have a smart team, things will work out. At that point, you have to keep iterating until you get something that customers really want.”
Part 2: Machine Learning Applications
Q: How has ML helped you improve the data collection and analysis process?
Yash: “We wanted our sensor to test for contamination in real-time. The real advantage of our product is giving brewers further insights from this data. ML is crucial to getting a high accuracy rate and analyzing that data quickly to give brewers actionable insights.”
Q: What is the advantage your tech has over industry alternatives?
Yash: “Our competitors are mostly providing expensive hardware for biological tests to companies. They have to use the hardware themselves and it gives a binary answer (contaminated or not). Software isn’t really involved. We wanted to abstract the hardware from the user and provide a platform to track much more than just contamination.”
Q: Why is it difficult to create an ML model in the biochem space?
Yash: “The reason it’s hard to develop this model in the biochem space is that we’re applying ML on experimental data, not just data you can buy. Most other tech companies can buy data from other sources but we have to produce it ourselves and analyze it on top of that. We had started at 70–80% accuracy and we’re constantly working to improve it.”
Part 3: Growth
Q: Have you talked to different breweries about your company?
Yash: “When we first started talking to breweries, we described our product and asked if it was a good idea. People liked the product but we didn’t have a price to give them. We should have done more interviews understanding their problems first and then build out our product. That’s when we changed our strategy. This helped us find different applications for our sensor outside of contamination that breweries wanted to analyze.”
Q: What is the process of setting up a PoC with a brewery?
Yash: “We reached out to 5–6 breweries and got 4 responses. Our manufacturing potential was not up to scale at that point so we could only partner with one. The pilot will last 3–4 months. After we generate enough data, we will start approaching larger brewers. Once that works, we can scale up our manufacturing process.”
If you’d like to personally get in touch with Yash, feel free to reach out to me at firstname.lastname@example.org and I can help set up an introduction.
The FreeV Spotlight and article were written and edited by Pratik Bhat, from the FreeV Internal Team.