QUICK SUMMARY
The FEI25 Collective Intelligence Startup Competition featured twelve innovative startups leveraging AI with human intelligence to solve critical business challenges. Chelsea Trengrove from Neoclease won the first prize of $25,000, while Roger Dias from Xtory took second place with $5,000. The competition showcased how collective intelligence can transform industries from healthcare to finance, with each founder demonstrating unique approaches to combining human expertise with AI capabilities.
KEY QUOTES
- “Imagine if there is a solution. An AI coworker that could be there side by side with clinicians 24/7 consistently providing high level intelligent recommendations and suggestions for all of these decisions that they make.” – Roger Dias, Xtory
- “Neoclease creates custom gene editors that are miniaturized and targeted to an individual gene. We’ve built a generative AI model and in silico evaluation pipeline that’s trained on millions of proteins that cut the DNA.” – Chelsea Trengrove, Neoclease
- “We have a couple of moats. One is bringing these organizations together to create the library… At the core of our technology is the library that we’re developing. It’s not the model, it’s the data underneath the model.” – Jason Radford, The Ostrea Cultura Company
- “This is collective intelligence, a system that scales with you. The more you use it, the smarter it becomes.” – Rose Huang, Accelidea
- “We are uniquely placed in that traditionally this is not a new problem. And traditionally, everyone in the market solving it is a consulting firm. And the way they solve it is they throw humans at it.” – Anushka Singh, Kathalyst
FULL SESSION SUMMARY
Chelsea Trengrove, Neoclease (First Prize Winner)
PITCH SUMMARY
Chelsea Trengrove presented Neoclease, a company creating custom gene editors that are miniaturized and targeted to individual genes. Their platform uses generative AI trained on millions of DNA-cutting proteins to output hundreds of thousands of potential editors for a single gene overnight. Computational biologists then rank the top 1% of these editors to determine which proceed to wet lab validation. This approach streamlines the discovery process for new drug candidates from two years to under six months. Their first program targets LARC2, a gene implicated in Parkinson’s disease affecting over 10 million patients worldwide. The business model involves licensing these bespoke editors to corporations or co-developing them as therapies.
Q&A SUMMARY
When asked about the competitive landscape, Chelsea explained that while many companies use a “one-size-fits-all” approach with proteins like Cas9, Neoclease uniquely optimizes each therapy for a single target. She highlighted partnerships with Amazon and Nvidia for resource support, the Michael J. Fox Foundation’s Light Consortium for LARC2 expertise, and wet lab space secured through golden tickets at BioLabs and Lab Central. Regarding regulatory challenges, she noted that the FDA has recently issued guidance on platform approaches to gene editing. Chelsea explained that their $5 million seed funding would progress their LARC2 candidate to IND-enabling studies and develop other gene-specific candidates for a mini pipeline.
Roger Dias, Xtory (Second Prize Winner)
PITCH SUMMARY
Roger Dias presented Xtory, an AI platform augmenting clinicians’ intelligence for better decision-making in emergency departments. He described emergency departments as complex environments where clinicians must make life-and-death decisions every minute, with both accuracy and speed being paramount. Suboptimal decision-making leads to preventable medical errors, inefficient workflows, and high healthcare costs. Xtory offers an AI coworker that provides intelligent recommendations 24/7, improving patient care while helping hospitals increase revenue and decrease unnecessary costs. Their proprietary technology leverages a hybrid AI architecture to transform raw data from electronic health records into valuable insights for clinicians.
Q&A SUMMARY
When questioned about the agentic nature of their application in high-stakes ER decisions, Roger clarified that their system always keeps humans in the loop for clinical recommendations. He explained that while they use agentic systems for data extraction and harmonization, clinicians always make the final decisions. Addressing concerns about real-time data processing, Roger detailed that their on-premises deployments in hospital servers can extract a patient’s 10-year history in one minute and process LLM responses in 3-4 seconds. He also mentioned their use of predictive neural networks for precision medicine applications, such as predicting sepsis, ICU needs, and mortality risks.
Anushka Singh, Kathalyst
PITCH SUMMARY
Anushka Singh presented Kathalyst, which addresses the critical problem of legacy COBOL code that powers essential systems like banking, insurance, and government services. With 800 billion lines of COBOL code running worldwide and a diminishing pool of COBOL engineers (average age 55+), these systems face significant maintenance challenges. COBOL AI uses machine learning to automatically convert legacy code to modern languages, reducing conversion time from years to weeks and cutting costs by 90%. Their solution has already been validated with pilot customers, including a Fortune 500 bank.
Q&A SUMMARY
When asked about the market opportunity, Anushka revealed a total addressable market of $750 billion that’s growing 10-20% annually. She explained that while their product currently focuses on COBOL, it can work with any software language, with plans to expand to Visual Basic, .NET, C/C++, and others. Regarding their competitive advantage, she noted that traditional solutions come from consulting firms that “throw humans at the problem,” whereas COBOL AI offers a scalable product solution in a space where domain knowledge is typically locked within consulting firms without motivation to build products.
Jason Radford, The Ostrea Cultura Company
PITCH SUMMARY
Jason Radford presented Ostrea Cultura, a company focused on cleaning misinformation from digital streams. He explained that misinformation destroys social connections, leading to real-world consequences like genocide and political instability, while causing platforms to face user disengagement and substantial fines. Their solution combines expert knowledge with generative AI to build models that detect individual claims and help moderators make decisions. They’re working with partners like Community Notes to assess experts, an online social media background check company to evaluate their model, and GGWP for moderator-facing services. Their target market is “buy not build” companies (Series B to Series D) that need this capability but lack resources to develop it in-house.
Q&A SUMMARY
When challenged about the subjective nature of misinformation, Jason clarified that they focus on objectively false claims and well-established misinformation narratives. Regarding the business model tension (platforms benefiting from engagement-driving misinformation), he acknowledged the short-term engagement boost but emphasized that long-term user retention and advertiser relationships suffer. He explained their pricing model as per-API-call, noting that content volume continues to increase. Jason also detailed their approach to identifying misinformation through a combination of human expertise and AI, with humans making final determinations.
Arnell Milhouse, yc0n1c
PITCH SUMMARY
Arnell Milhouse presented yc0n1c, an agentic AI, blockchain, and human intelligence integration platform designed to disrupt traditional venture capital. He described their solution as reinventing the global VC ecosystem through merit-based capital allocation, where companies with traction receive funding based on their performance. The platform is tokenizing startups by putting them on blockchain, using their traction as a consensus mechanism to mint YCO tokens. Their collective intelligence diagram ensures AI decisions are reviewed by DAO validators before triggering smart contracts, with a vetting framework to ensure trustworthiness of all DAO members.
Q&A SUMMARY
When asked about their customer base, Arnell identified founders and investors as primary customers, but noted that their DAO also allows people to participate as validators without being founders or investors, creating what he calls the “DAO Lancer economy.” Addressing the tension between equitable funding and the risk-reward nature of venture capital, he explained that the platform embraces risk but focuses on helping founders solve risk assessment at early stages through their AI (which has its own personality and social media presence). Regarding traction measurement, Arnell mentioned they have about 1,000 different metrics ranked from 0-1 in terms of importance for minting YCO tokens.
Sesha Kadakia, Tangify
PITCH SUMMARY
Sasha Kadakia presented Tangify, an AI-powered solution addressing the outdated process of documenting and capturing intellectual property. She highlighted that while innovation speed has increased, IP documentation remains stuck in decades-old processes, with engineers typically writing brief descriptions and throwing them “over the wall” to IP teams. Tangify gives inventors “IP superpowers” by quickly identifying inventions, sifting and synthesizing them, and instantly generating disclosures for seamless handoffs to IP colleagues. What traditionally takes 6-12 months can now be done in minutes, with customers “banging down their door” since their beta launch in January.
Q&A SUMMARY
Regarding their business model, Sasha explained that for smaller companies, they use a per-user license model, while larger organizations pay a fixed platform fee based on company size, which includes access to a certain number of seats. On data governance, she emphasized that all data remains private and owned by individual companies, with a “burn button” allowing immediate removal from the system. Sasha noted that their entry point is typically through corporate legal departments or IP teams, with their end goal being complete integration into product and engineering workflows, eventually moving toward continuous monitoring that proactively alerts users.
David Hojah, Parrots
PITCH SUMMARY
David Hojah presented Parrots, a platform helping patients with neurological disorders. Motivated by his uncle’s Alzheimer’s diagnosis, David and his team (including MDs, PhDs, and neurologists) built Poly, a platform for early detection of signs and symptoms of dementia and Alzheimer’s. Their first product is a SaaS solution where patients use a smartwatch and play games on an iPad, allowing healthcare providers to conduct cognitive assessments. Their second-generation product will be a headband that reads brain activity. The business model involves working with healthcare providers, with expansion plans from nursing homes to hospitals and insurance coverage. They have four issued patents, two pending, and are HIPAA compliant.
Q&A SUMMARY
When asked about their benchmarking against current diagnostics, David explained they focus on early-stage Alzheimer’s patients, with plans to help those not yet officially diagnosed. He emphasized their comprehensive approach, combining brain activity monitoring, speech analysis, cognitive behaviors, and memory loss assessment with machine learning, distinguishing them from competitors who typically focus on just one aspect. Regarding efficacy, David noted their system has reduced healthcare provider demand by 30% (from 100 hours to 70 hours) and helps track conditions before they reach crisis points requiring emergency room visits.
Rose Huang, Accelidea
PITCH SUMMARY
Rose Huang presented Accelidea, addressing the increasing time it takes to bring medical devices to market (from 34 months in 2005 to 47 months today) due to documentation challenges. Having worked in the industry for over a decade, Rose built Accelidea with patent-pending technology that automates up to 80% of documentation work across product management, engineering, regulatory, and quality teams. The system accesses public and proprietary databases to provide product codes and regulations within minutes, then learns and updates all downstream documents instantly as specifications are reviewed. The market for medical device companies is approximately $13 billion.
Q&A SUMMARY
When asked about regulatory changes affecting their business, Rose noted that the FDA recently announced plans to use AI internally, which she sees as positive. She emphasized that FDA oversight remains essential for medical device safety. Regarding data security for proprietary information, Rose explained that they keep customer data separate, have agreements about training data usage, and clearly indicate which data is human-validated versus AI-extracted. She highlighted their traceability feature (currently in the roadmap) as a key selling point, providing audit trails that show exactly what decisions led to specific testing requirements.
Liz Graham, Ada IQ
PITCH SUMMARY
Liz Graham presented Ada IQ as “Moneyball for consumer products,” taking the guesswork out of developing products that sell. She highlighted that 40% of new products fail or underperform when they hit the market, not due to lack of data but because teams are overwhelmed with information and rely on subjective gut feel. Ada IQ helps teams integrate disparate data sets, identify attributes that matter, and provide instant access to concept evaluation. With over a million dollars in grant funding and a dozen published academic papers supporting their technology, they help teams deepen customer focus, accelerate product development cycles, and build products that sell quickly.
Q&A SUMMARY
When asked about the competitive landscape, Liz explained that while there are many companies focused on consumer insights and others like Adobe Illustrator offering generative design platforms, Ada IQ uniquely fuses these approaches. Their patented models use a deep multimodal design evaluation approach that combines sales data, customer reviews, and visual images to identify attributes that matter to consumers.
Dean (Dien) Hu, Caid
PITCH SUMMARY
Dean presented Caid (Computer Aided Design with Applied AI Lab), addressing the challenges of rapid prototyping which requires significant investment, complexity, and time. Their solution focuses on converting sketches and images to CAD models without requiring natural language prompting, which Dean identified as a fundamental mistake in current approaches. Caid allows real-time editing of models for professional users and features self-refining capabilities. Their pilot studies demonstrated at least 10x speed improvements compared to traditional methods. Dean outlined a timeline leading to what he described as “the world’s first AI 3D printer” and emphasized their in-house development of sketch-to-CAD technology.
Q&A SUMMARY
When questioned about their early-stage status and user base, Dean emphasized the importance of starting small to make a core group of users happy before expanding. He clarified that their input method relies on sketches and images rather than text prompts, which he believes is superior to natural language interfaces for 3D design. Dean distinguished Caid from Autodesk by explaining that Autodesk focuses on generative design and engineering processes, while Caid specializes in sketch/image-to-CAD conversion. He suggested their technology could democratize 3D design for non-specialists, making the design process more experimental rather than requiring specialized training. Potential applications include 3D printed food, car parts, and robot components.
Shawn Harris, Coworked
PITCH SUMMARY
Shawn Harris presented Coworked, addressing the labor crisis in project management where 25 million more project managers will be needed by 2030. Their solution, Harmony, is an agentic AI project manager capable of handling 70% of typical project management work. He explained that project managers operate in the white space between people, processes, and systems, and Harmony allows human project leaders to focus on strategy and business objectives while the AI handles routine activities like setting up projects, tracking status reports, and coordinating teams. Their business model charges based on active projects under management rather than per user. Shawn mentioned they have already shipped a commercial product, have a team of 10 engineers, and have secured MSAs with two Fortune 500 enterprises.
Q&A SUMMARY
When asked about the challenges of getting new software approved by corporate IT departments, Shawn confirmed they’ve successfully navigated this process with their enterprise clients. He explained that Harmony integrates into existing environments as a resource with its own email address and calendar access, functioning like a remote colleague rather than introducing a new UI. The system integrates with various tools including Workday, Salesforce, and project management platforms, using them as tools while working alongside human team members. He distinguished Coworked from established project management platforms like Asana and JIRA by noting that these companies would face an innovator’s dilemma in trying to build something like Harmony, as it would disrupt their own business models. He explained that Harmony comes pre-loaded with best practices for project management and learns enterprise-specific processes over time.
Matt Bowring, Gigabug Computer
PITCH SUMMARY
Matt Bowring, founder and sole engineer at Gigabug Computer, presented an innovative analog computing solution addressing the exponential growth in energy required for AI data centers. As transistors approach their thermal efficiency limits, Matt proposed a novel computing paradigm for optimization problems like navigation, budgeting, resource allocation, and scheduling. By designing analog circuits that naturally minimize energy potentials, his solution allows physics to solve problems more efficiently than traditional computing methods. Matt showcased a prototype circuit board he designed and manufactured, claiming it could be orders of magnitude more energy efficient than GPUs, TPUs, and CPUs for specific optimization problems.
Q&A SUMMARY
When asked about his target customers, Matt identified organizations with compute-intensive workloads, including big tech companies like Walmart and Amazon that deal with complex scheduling and navigation problems. He also mentioned government interest from DARPA, the Department of Defense, and the Department of Energy. Regarding competition with established players like Nvidia, Matt explained that his approach differs fundamentally as an analog-based computer rather than a digital architecture. His business strategy involves starting small by offering low-cost development kits to allow people to explore programming these devices. For his product roadmap, Matt plans to design a next-generation architecture within a year and send it to a foundry using the latest technology.
KEY TAKEAWAYS
- Collective Intelligence Amplifies Innovation: The winning startups demonstrated how combining human expertise with AI capabilities creates more powerful solutions than either could achieve alone, whether through gene editing (Neoclease), clinical decision support (Xttory), or legacy code conversion (Kathalyst).
- Domain Expertise Remains Critical: Despite advances in AI, deep domain knowledge continues to be a competitive advantage, as seen in Neoclease’s computational biologists selecting the most promising gene editors and Kathalyst’s understanding of legacy systems.
- Human-in-the-Loop Design Prevails: The most compelling solutions maintained humans in critical decision loops while automating routine tasks, as demonstrated by Xtory’s approach to clinical recommendations and Ostrea Cultura’s misinformation detection.
- Democratization of Access: Several startups focused on democratizing access to previously exclusive domains, such as yc0n1c’s approach to venture capital and Tangify’s streamlining of IP processes.
- Regulatory Navigation is Key: Successful startups showed awareness of regulatory landscapes and built compliance into their solutions, as seen Accelidea’s FDA documentation approach and Neoclease’s attention to FDA guidance on gene editing.
DELIVERY ON EVENT FOCUS: Aligning Innovation with Business Strategy
The competition showcased startups that directly aligned innovative technologies with clear business strategies and market needs. Each finalist demonstrated how their solution addresses specific pain points while creating sustainable business models. For example, Neoclease’s approach to gene editing not only advances scientific capabilities but also creates a clear licensing and co-development business model. Xtory showed how their clinical decision support tool improves patient outcomes while simultaneously helping hospitals increase revenue and decrease costs. Kathalyst identified a critical business need (maintaining legacy systems) and developed a scalable solution with a massive addressable market. This alignment of cutting-edge technology with practical business applications exemplifies how innovation can be strategically deployed to create value.
DELIVERY ON EVENT THEME: Harvesting Innovation and Sowing the Seeds of Future Growth
The competition embodied the theme of harvesting innovation by showcasing mature applications of AI and other technologies ready for immediate implementation, while simultaneously sowing seeds for future growth through forward-looking approaches. Startups like Tangify are harvesting the benefits of AI to solve current IP documentation challenges while sowing seeds for future continuous monitoring capabilities. Neoclease is applying existing AI capabilities to current gene editing needs while developing a platform that could transform multiple disease treatments in the future. yc0n1c is harvesting blockchain and AI technologies to reimagine venture capital today while sowing seeds for a new “DAO Lancer economy” that could create opportunities for millions displaced by automation. This dual focus on immediate application and future potential demonstrates how innovation can be both harvested for current value and cultivated for long-term growth.
ACTION ITEMS FOR INNOVATION EXPERTS & CORPORATE CHANGEMAKERS
- Audit Your AI Integration Strategy: Evaluate how your organization combines human expertise with AI capabilities, ensuring you’re leveraging the strengths of both rather than simply automating existing processes.
- Identify Legacy System Vulnerabilities: Assess your organization’s dependence on legacy systems and develop modernization strategies that preserve institutional knowledge while enabling future innovation.
- Implement Human-in-the-Loop Design Principles: Review your AI implementations to ensure appropriate human oversight at critical decision points while automating routine tasks.
- Explore Decentralized Innovation Models: Consider how blockchain and DAO structures might enable more equitable and efficient innovation funding and collaboration within your organization or industry.
- Develop Comprehensive IP Capture Processes: Implement systems to identify and document intellectual property throughout the innovation process, not just at formal disclosure points.
- Partner with Startups Strategically: Identify startups whose collective intelligence approaches could complement your organization’s capabilities, focusing on those that combine domain expertise with technological innovation.
- Build Regulatory Navigation Capabilities: Develop expertise in how emerging technologies interact with regulatory frameworks in your industry, positioning your organization to innovate within compliance boundaries.
- Create Cross-Functional Innovation Teams: Form teams that combine technical, domain, and business expertise to ensure innovations address real needs and have viable paths to market.