[Disclosure: AgFunderNews’ parent company AgFunder is an investor in Brightseed.]
Brightseed—a startup on a mission to illuminate the ‘dark matter of nutrition’ hidden in the natural world—has launched Hummingbird, an agentic AI product sitting on top of its Forager bioactives discovery platform to enable partners to move from discovery to development in record time.
Founded by Lee Chae PhD, Sofia Elizondo, and Jim Flatt PhD, in 2017, Brightseed has built a vast database of 21 million plant-based compounds associated with biological targets. Forager can predict which ones might confer specific health benefits and help identify the optimal way to source or produce them at commercial scale.
The platform was first used by Brightseed to develop its own bioactives, two of which (NCT and NFT) it has now commercialized as gut and metabolic health ingredients. It was then opened up to third parties, enabling leading food and nutrition companies to benefit from the same cutting-edge tools.
Hummingbird takes this a step further by layering on agentic AI to close the bottlenecks that were preventing Brightseed from scaling the platform.
‘Let’s make AI the commercial engine’
Chae tells AgFunderNews: “We’ve partnered with a number of industry players in nutrition, health and wellness and supplements over the years. But it was a hard business model to scale as it [the workflow] was manual, so our team would [use Forager] to spend weeks doing the research, producing reports, and then our customers might take weeks to digest the information and then come back with more questions.
“With Hummingbird, scientists, innovation leads, product developers and product marketers [at Brightseed’s customers] can use the platform, ask those same questions [directly] using natural language and get answers back in seconds versus days or weeks, and do this on a continual basis.
“One of our customers came back to us and said that through Hummingbird, one of its PhD scientists just did in two weeks with our platform what it used to take six months to achieve.”
He adds: “We’ve already invested years building, validating and commercializing insights from our AI so the whole idea was let’s make the AI our commercial engine. The markets we serve aren’t quite used to having pharma-type discovery platforms, so we want to make it easier for them to develop and commercialize differentiated health products that address consumer needs by bringing the AI out from behind the curtain.”
From discovery to development
While everyone is using generic AI tools for research these days, many food and nutrition companies surveyed by Brightseed said they still felt like they were “fishing in the sea” when it came to discovery and “mixing and hoping” when it came to development, says Chae.
“Take [a hypothetical client called] Kimberly, a product developer for a nutraceutical company. She has a mandate to develop, say, 10 new product concepts every quarter to bring to her steering committee. Today she listens to podcasts, reads blogs, goes to conferences, calls her suppliers, reads the scientific literature.”
With Hummingbird, he says, Kimberly can “say something like: I want to develop a new product to help women aged 35-55 fall asleep.”
Hummingbird then interrogates Brightseed’s proprietary Forager platform and starts to return recommendations on the types of compounds Kimberly can use, the biomarkers to test for, the rationale around pathways, and credible scientific evidence from the literature that supports it. Then it starts building a product brief.
According to Chae: “It also gives dosage information so formulators can say, okay, because of that, we know it will work in these product formats, say in a beverage, tablet or gummy. It’s going way beyond just basic scientific information.”
It can also help commercial teams explore potential claims they could make based on the science, and whether more research is needed, compare the product concept to what’s already on the market, and address IP issues “so that [the] legal [team] and the business owner can say, Yeah, this is a unique opportunity for us to really own this formulation.”
Evaluating scale-up options
For certain bioactives identified by Hummingbird, there may be well-established botanical supply chains, says Chae. For others, Hummingbird can help identify the optimal way to source or produce them at commercial scale.
This might mean extracting them from a plant, but may involve biomanufacturing, he says. In this case, Hummingbird can tap into Forager’s understanding of the metabolic pathways that plants use to produce the bioactive compound in question and then use those pathways as a blueprint for producing it via precision fermentation, for example.
“We’re building out a scale up agent that will allow you to assess different routes to production, be it a natural extract, be it precision fermentation, be it biotransformation, or chemical synthesis.”
Mining natural compounds for GLP-1 activity
Right now, says Chae, some customers are using Hummingbird—which has been in test mode since February—to explore bioactives that might mimic GLP-1 or trigger its production in the body.
“We already have all of those functional inference models encoded in the platform and the chemical space encoded to be able to query against those models and allow potential hits. And we already have the biology knowledge in the system around those pathways. It’s a life science model that connects natural sources and their chemicals to the human body and their health benefits.”
Why generic AI tools fall short
While potential customers now have a plethora of AI tools at their disposal, Claude, Gemini, and ChatGPT et al will not deliver the kind of actionable insights they can get on Hummingbird, which is pulling from a proprietary dataset Brightseed has been building for years, says Chae.
“We just got visited by the global executive team of one of the top companies in its field. They saw a demonstration of the platform, and the CEO went home that night and asked the same exact same questions [he had asked Hummingbird] to all the big [LLM] models.”
The difference was stark, claims Chae. “He said your system provided more answers, answers that the other models didn’t deliver, with more detail, more explanation. He then asked his team which insights would be more actionable and they said the recommendations coming from Brightseed’s system were way more actionable.”
New agents rolling out quarterly
Agents associated with different tasks will be launched rapidly in the coming months and years.
“The ‘white space’ agent is in alpha [testing] already, and we’re going to release it to beta soon so our customers can play with it,” says Chae.
“A claims agent is coming out in Q2, and we have also have a formulation agent, a scale up agent, and a market agent. We have a whole roadmap behind this; our goal is to be able to deploy a new agent at least once a quarter.”
Pricing models and progress to date
There are three pricing models. Customers can pay for a subscription to the platform, they can pay for data generation using their own proprietary data, and they can pay for Brightseed’s (human) experts to help them formulate ideas, interpret data, plan validation studies, and find and manage CROs.
According to Chae: “A lot of companies have libraries of botanical extracts or microbial strains that they’ve developed. We can generate the data on that in terms of what chemistries are present and how they apply to human health.”
It’s early days, he concedes, but Brightseed has already onboarded two customers since February and is now “contracting with a handful more that should close out in Q2. So right off the bat, we’re almost at our year’s target for customers, just in the first four months of Hummingbird being out in the market.”
Further reading:
Brightseed addresses gut barrier function with bioactives backed by new human clinical data
Brightseed and Haleon partner to harness AI for small molecule discovery
Manitoba Harvest and Brightseed target gut barrier function with bioactive fiber launch
Meet the founder: Brightseed’s Dr. Jim Flatt illuminates the ‘dark matter of nutrition’
