IA Seed Bets Vertical AI Can Slash 10-Year Drug Approvals to Months
Ai4
Vertical AI startups can eliminate traditional biotech binary risk by pairing immediate diagnostic cash flow with automated discovery pipelines that bypass legacy wet-lab testing.
## Background
Published on YouTube by Ai4 on August 28, 2025, this presentation features Ed Dua of IA Seed Ventures and Dr. Christin Glorioso of NeuroAge Therapeutics. According to the description, the session explores how narrow, AI-first vertical slices unlock rapid startup growth in healthcare, spotlighting NeuroAge's approach to Alzheimer's diagnostics and therapeutics.
## TL;DR
Venture capitalist Ed Dua and NeuroAge Therapeutics CEO Dr. Christin Glorioso discuss how specialized vertical AI applications are disrupting healthcare and legacy industries. Glorioso details how NeuroAge couples immediate clinical brain-aging diagnostics with causal and generative AI to tackle historically undruggable therapeutic targets for Alzheimer's. Dua outlines how portfolio startups tackling in silico simulation, lab robotics, and EHR patient matching can compress traditional drug development timelines from a decade to months. The presentation concludes by showing how full-stack vertical AI disrupts disparate fields beyond medicine, from aerospace metal manufacturing to quadruped robotic guide dogs.
## Atmosphere & style
The presentation carries an upbeat, collegial investor-showcase vibe, with Ed Dua enthusiastically acting as an encouraging advocate for technical founders. Dr. Glorioso brings grounded scientific authority and heartfelt personal motivation to the stage, moving smoothly from her family's history with Alzheimer's to complex causal genomics. The dynamic feels crisp, optimistic, and pragmatic, capturing the fast-paced excitement of early-stage venture building applied to stubborn, high-stakes real-world problems.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=4s) The core investment thesis of IA Seed Ventures focuses on 'vertical AI slices'—narrow, AI-first entry points that radically disrupt legacy workflows across diverse industries.
- [09:18](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=558s) Early-stage venture opportunities in AI are drawing comparisons to the early SaaS era, where lower cost structures and frictionless models exponentially expanded addressable markets beyond legacy spend.
- [09:18](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=558s) The enterprise landscape is dividing into three categories: legacy incumbents attempting to retrofit AI, full-stack vertical startups displacing incumbents, and pure AI natives creating entirely new categories.
- [02:57](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=177s) Biological brain aging represents approximately 40% of Alzheimer's risk, and maintaining a biological brain age five years younger correlates with a sixfold reduction in disease risk.
- [02:57](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=177s) NeuroAge utilizes a dual commercial-research model that generates immediate clinical test revenue from brain age assessments while accumulating multi-omic data for long-term drug discovery.
- [05:37](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=337s) Combining causal inference techniques like Mendelian randomization with generative AI drug design allows startups to target the ~90% of disease-associated proteins traditionally deemed undruggable.
- [09:18](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=558s) Startups targeting individual bottlenecks in the 7-to-10-year FDA drug development pipeline (such as in silico simulation, lab robotics, and instant trial matching) could dramatically shorten approval timelines.
- [13:16](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=796s) Full-stack AI manufacturing startups can displace legacy industrial suppliers by integrating automation across robotics, operations, finance, and customer service.
- [13:16](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=796s) Embodied AI robotics companies, such as autonomous guide dog developers, can disrupt high-cost services (e.g., $70,000 guide dog training) while building proprietary datasets from spatial navigation and human interaction.
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=4s) Vertical AI Slices: An Era of Ubiquitous Disruption
Ed Dua introduces himself as the manager of a Berkeley-based early-stage seed fund, IA Seed Ventures, reflecting on more than 15 years of early-stage investing. He highlights that across his investing career—which includes five seed-stage investments that achieved IPOs alongside numerous high-return exits—he has "never been more excited in terms of the opportunities." The driving thesis of the fund centers around what he terms "vertical AI slices."
```mermaid
flowchart TD
accTitle: Vertical AI Drug Discovery and Clinical Pipeline
subgraph Discovery ["Target Discovery & Design"]
A["Multi-omic Data & Causal AI<br/>(NeuroAge)"] --> B["Generative Protein Design<br/>(Novel & Undruggable Targets)"]
end
subgraph Testing ["Preclinical In Silico Validation"]
B --> C["In Silico Simulation<br/>(Angstrom AI Wet-Lab Replacement)"]
C --> D["Robotic Lab Automation<br/>(Xeon Systems)"]
end
subgraph Trials ["Clinical Translation & Validation"]
D --> E["AI-Assisted Patient Matching<br/>(HealthKey Database Search)"]
E --> F["Shortened FDA Timeline<br/>(Target: Months vs. 7-10 Years)"]
end
```
Dua argues that artificial intelligence will not merely enhance software, but will fundamentally reshape every single sector across an immense gamut of industries—from legacy industrial manufacturing like sheet metal cutting to high-tech robotics aiding the visually impaired. To illustrate how this paradigm plays out in healthcare, Dua introduces Dr. Christin Glorioso, CEO and co-founder of portfolio company NeuroAge Therapeutics, jokingly offering to act as her "cheerleader" because technical founders frequently understate their achievements.
- Ed Dua has invested in early-stage seed ventures for over 15 years, securing 5 seed-to-IPO exits.
- The investment thesis focuses on narrow, highly specialized AI-first entry points ("slices") that transform legacy workflows.
---
[02:57](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=177s) NeuroAge Therapeutics and Tackling Brain Aging
Dr. Christin Glorioso outlines her background as an MIT-trained physician and neuroscientist, as well as the former head of AI at a Stanford spinout, before founding NeuroAge Therapeutics roughly three years ago. Her work is driven by deep personal experience: watching her grandfather care for her grandmother throughout a grueling seven-year battle with Alzheimer's disease. During the creation of NeuroAge, Glorioso discovered that she carries one copy of the *ApoE4* allele—a genetic risk factor present in roughly 20% of the population that doubles or triples an individual's risk of developing Alzheimer's in their 70s or 80s.
Rather than treating late-stage symptoms with therapies that often carry severe side effects, NeuroAge focuses on biological brain aging—the single largest risk factor for Alzheimer's, accounting for approximately 40% of overall disease risk. Glorioso notes that achieving a biological brain age five years younger than one's chronological age correlates with a sixfold reduction in future Alzheimer's risk.
To capitalize on this insight, NeuroAge operates as a "tech bio" company generating immediate commercial revenue through a multi-modal brain aging assessment that integrates brain MRIs, cognitive evaluations, and proprietary blood biomarkers. This assessment provides patients with actionable lifestyle interventions while feeding a unique, multi-omic longitudinal dataset to uncover causal genetic pathways and develop preventative longevity drugs alongside pharmaceutical partners.
- Carrying one *ApoE4* allele elevates Alzheimer's risk by 2x to 3x in elderly years.
- Biological brain aging represents ~40% of Alzheimer's risk; being biologically 5 years younger reduces Alzheimer's likelihood by 6x.
- NeuroAge couples clinical test revenue with long-term drug target discovery.
---
[05:37](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=337s) Unlocking Undruggable Targets with Causal and Generative AI
Dua emphasizes that NeuroAge exemplifies the ideal seed investment: it avoids the purely speculative binary risk of traditional biotech by securing early testing revenue while cultivating therapeutic assets. Dua notes that current Alzheimer's treatments are "really awful" in terms of side effects, making proactive longevity interventions essential.
Glorioso elaborates on how NeuroAge deploys causal AI techniques—such as Mendelian randomization and causal inference—to disentangle correlation from causation across aging pathways, recognizing that "not all Alzheimer's is the same" and tailoring treatments to individual biological subtypes.
Addressing a key bottleneck in pharmacology, Glorioso explains that roughly 90% of disease-associated proteins have historically been classified as "undruggable." Through partnerships with generative AI drug design platforms, NeuroAge is designing bespoke, de novo therapeutic molecules tailored to previously inaccessible biological targets, effectively aiming to "10x all of the drugs that are possible for diseases."
- 90% of functional disease proteins are currently considered undruggable using conventional discovery pipelines.
- NeuroAge employs Mendelian randomization to pinpoint precise mechanistic drivers of accelerated cognitive decline.
---
[09:18](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=558s) Transforming the FDA Pipeline: In Silico Testing and Trial Acceleration
Dua draws a historical parallel between modern vertical AI and the early transition from on-premises enterprise software to Software-as-a-Service (SaaS). Skeptics originally argued that SaaS market caps would remain constrained by legacy on-prem IT spend, failing to foresee that dramatically lower costs and frictionless delivery would expand the total addressable market exponentially. Dua categorizes the emerging enterprise landscape into three distinct buckets:
| Category | Description | Primary Challenge / Advantage |
|:--- |:--- |:--- |
| **Legacy Incumbents** | Established enterprises trying to retrofit AI | Massive organizational inertia; high risk of disruption |
| **Full-Stack Vertical AI** | Startups rebuilding entire operational stacks | Dramatically lower cost structures; unseats legacy leaders |
| **Pure AI Natives** | Companies executing models impossible without AI | Creates entirely new markets and capabilities |
Because traditional drug development spans 7 to 10 years to navigate the FDA pipeline, IA Seed Ventures targets specific friction points along that continuum:
- **Angstrom AI:** Replaces slow, capital-intensive wet-lab assays with high-throughput *in silico* computer simulations during preclinical validation.
- **Xeon Systems:** Integrates laboratory robotics with machine learning to automate and accelerate wet-lab testing throughput.
- **HealthKey:** Leverages AI to parse complex hospital electronic health records and instantly match eligible patient cohorts to clinical trial exclusion/inclusion criteria in minutes—a process that historically consumed months or years.
Dua envisions an eventual future where compressing discovery, preclinical validation, and clinical matching could theoretically reduce regulatory drug approval turnaround from a decade down to a matter of months.
---
[13:16](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=796s) Full-Stack AI Startups: SpaceX Suppliers to Robotic Guide Dogs
Stepping outside healthcare, Dua highlights the extreme diversity of vertical AI implementations through two contrasting portfolio examples:
* **Disrupting Legacy Metal Fabrication:** Dua profiles a founder who spent six years at an industrial supplier slicing large aluminum plates for aerospace customers such as SpaceX. Rather than attempting to patch legacy workflows, the founder is building a full-stack, AI-native manufacturing company from the ground up—integrating AI across operations, finance, customer service, sales, and automated robotics. By operating with a dramatically leaner cost structure, the startup aims to directly displace incumbent suppliers.
* **Talis Robotics (AI Seeing-Eye Dogs):** Founded by two roboticists who earned PhDs at age 21, Talis is developing an embodied AI robotics platform. Their initial flagship application deploys quadruped robot dogs as autonomous seeing-eye companions for the visually impaired. With guide dog training currently costing upwards of $70,000 per animal, Talis provides an accessible alternative while aggregating a valuable proprietary data lake—combining real-time autonomous physical navigation data with rich interaction data between the user and the robot.
Dua concludes by noting that whether a startup is cutting aerospace metal or training robotic seeing-eye dogs, narrow vertical slices that build proprietary data moats represent the vanguard of early-stage AI venture creation, before opening the floor to audience questions.
- Professional seeing-eye dog training currently costs approximately $70,000 per dog.
- Talis Robotics creates dual-layer data lakes capturing both spatial navigation and human-robot interaction patterns.
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=4s) never been more excited in terms of the opportunities — Ed Dua
> [00:04](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=4s) cheerleader — Ed Dua
> [05:37](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=337s) really awful — Ed Dua
> [05:37](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=337s) not all Alzheimer's is the same — Dr. Christin Glorioso
> [05:37](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=337s) 10x all of the drugs that are possible for diseases — Dr. Christin Glorioso
## Mentioned
Ed Dua, IA Seed Ventures, Dr. Christin Glorioso, NeuroAge Therapeutics, MIT, Stanford, ApoE4, FDA, Angstrom AI, Xeon Systems, HealthKey, SpaceX, Talis Robotics
## Video description
Ed Dua, General Partner, IA Seed Ventures
Christin Glorioso, Co-Founder & CEO, NeuroAge Therapeutics
Abstracts
Where Will AI Impact Healthcare? Everywhere.
AI is fueling the creation of thousands of healthtech startups—some disrupting massive incumbents, others building businesses that simply weren’t possible before. In this session, we’ll dive into the concept of AI vertical slices—narrow, AI-first entry points into the healthcare system that can unlock rapid growth. We’ll also spotlight one of IA Seed Ventures’ most promising early-stage investments: NeuroAge Therapeutics, a startup using AI to develop groundbreaking Alzheimer's diagnostics and therapeutics.
IA Seed 押注垂直 AI 可将 10 年药物审批缩短至数月
垂直 AI 初创公司通过将即时诊断现金流与绕过传统湿实验的自动化发现管线相结合，能够消除传统生物科技的二元风险。
风险投资人 Ed Dua 与 NeuroAge Therapeutics 首席执行官 Christin Glorioso 博士探讨了专业化垂直 AI 应用如何颠覆医疗及传统行业。Glorioso 详细介绍了 NeuroAge 如何将即时临床大脑衰老诊断与因果及生成式 AI 相结合，以攻克阿尔茨海默病历来难以成药的治疗靶点。Dua 阐述了其投资组合中专注计算机模拟、实验室机器人以及电子健康档案（EHR）患者匹配的初创公司，如何将传统的药物研发周期从十年压缩至数月。演讲最后展示了全栈垂直 AI 如何在医药之外颠覆不同领域，从航空航天金属制造延伸至四足机器导盲犬。
IA Seed Ventures 的核心投资策略聚焦于“垂直 AI 切片”（Vertical AI Slices），通过高度垂直的 AI 原生切入点重塑传统工作流，其市场扩张潜力被类比为早期的 SaaS 浪潮。 (0:04)
NeuroAge 专注于占阿尔茨海默病风险约 40% 的大脑生物学衰老，采用兼顾即时临床检测收入与多组学数据积累的双重商业模式，生物学脑龄年轻 5 岁对应着发病风险降低 6 倍。 (2:57)
传统药物研发中约 90% 的致病蛋白被归类为“不可成药”，NeuroAge 结合孟德尔随机化等因果推理技术与生成式 AI 蛋白质设计，致力于大幅拓宽针对神经疾病的可治疗靶点。 (5:37)
针对通常长达 7 至 10 年的 FDA 药物研发管线，IA Seed Ventures 布局了 Angstrom AI（计算机模拟测试）、Xeon Systems（实验室机器人）及 HealthKey（即时临床试验匹配）等切入关键瓶颈的初创企业。 (9:18)
垂直 AI 正向医疗之外拓展，包括以全流程自动化颠覆传统供应链的航空航天金属加工初创公司，以及开发四足导盲机器狗以替代单只培训成本达 7 万美元传统导盲犬的 Talis Robotics。 (13:16)
Reading summary
IA Seed Ventures manager Ed Dua argues that specialized "vertical AI slices"—startups rebuilding entire industry stacks rather than retrofitting existing software—are poised to transform traditional markets through lower cost structures and proprietary data moats. In healthcare, this model is illustrated by tech-bio companies combining immediate diagnostic testing revenue with causal and generative AI to treat biological brain aging and accelerate drug development pipelines.
Full-Stack Vertical AI Over Legacy Retrofits (0:04)
Dua draws a direct parallel between modern vertical AI and the historical shift toward Software-as-a-Service (SaaS), asserting that market observers routinely underestimate how dramatic cost reductions and frictionless deployment expand total addressable markets. In his taxonomy of enterprise transition, organizations fall into three categories: legacy incumbents struggling with organizational inertia as they attempt to retrofit AI, full-stack vertical AI startups rebuilding entire operational stacks from scratch, and AI-native firms pursuing business models that were previously impossible.
Dua argues that narrow vertical entry points allow early-stage ventures to disrupt incumbent workflows from the ground up. By controlling their entire operational footprint—from customer service and finance to physical execution—these startups aim to displace established players with substantially leaner cost structures rather than selling software add-ons to incumbents.
De-Risking Biotech with Diagnostic Revenue (2:57)
NeuroAge Therapeutics co-founder Dr. Christin Glorioso outlines a tech-bio structure designed to mitigate the binary clinical risks inherent in traditional early-stage biotechnology. Instead of focusing solely on symptomatic late-stage therapies with severe side effects, NeuroAge targets biological brain aging, which Glorioso identifies as representing approximately 40% of overall Alzheimer's disease risk. According to Glorioso, achieving a biological brain age five years younger than chronological age correlates with a sixfold reduction in Alzheimer's risk, while carrying a single ApoE4 allele—present in roughly 20% of the population—doubles or triples an individual's risk in their 70s or 80s.
To establish financial resilience, NeuroAge pairs long-term therapeutic discovery with immediate commercial revenue generated from a multi-modal brain aging assessment. This assessment integrates brain MRIs, cognitive evaluations, and proprietary blood biomarkers to deliver personalized lifestyle recommendations. Dua notes that upfront testing revenue funds ongoing operations while building an expanding longitudinal multi-omic dataset for therapeutic discovery.
Causal Inference and Generative Protein Design (5:37)
Addressing historical constraints in drug development, Glorioso highlights that approximately 90% of disease-associated functional proteins have traditionally been categorized as undruggable. To systematically identify actionable intervention points, NeuroAge applies causal AI methodologies, such as Mendelian randomization and causal inference, to disentangle correlation from causation across multi-omic datasets and account for biological subtypes of Alzheimer's.
Once causal disease drivers are isolated, NeuroAge collaborates with generative AI drug design platforms to engineer de novo molecules tailored to previously intractable targets. Glorioso claims this combined framework of causal target validation and generative design aims to expand the universe of druggable disease targets by tenfold.
Target Disruption Across the Clinical Pipeline (9:18)
Standard drug development spans seven to ten years to satisfy the FDA regulatory pipeline. Dua details several early-stage portfolio investments targeting specific friction points along this development timeline: Angstrom AI replaces slow, capital-intensive wet-lab assays with high-throughput in silico simulations during preclinical testing; Xeon Systems incorporates machine learning with robotics to automate laboratory testing; and HealthKey utilizes AI to parse electronic health records, matching patient cohorts to trial criteria in minutes instead of months or years.
Dua envisions a theoretical future where integrating in silico validation, automated laboratories, and algorithmic cohort recruitment compresses the multi-year regulatory timeline down to a matter of months, though this acceleration remains an ambitious industry forecast subject to regulatory validation.
Proprietary Data Moats in Physical Industries (13:16)
Dua emphasizes that vertical AI disruption extends beyond life sciences into physical industries where startups can capture unique data moats. In industrial manufacturing, an industry veteran is developing a full-stack, AI-native metal fabrication company supplying aerospace clients like SpaceX, integrating automation across sales, finance, operations, and robotics to undercut legacy plate-cutting suppliers.
Similarly, Talis Robotics is developing quadruped robot dogs as autonomous guide companions for the visually impaired, addressing traditional guide dog training costs that approximate $70,000 per animal. Dua explains that the platform simultaneously builds a proprietary dual-layer data lake comprising real-time autonomous navigation data and continuous human-robot interaction patterns, illustrating how vertical integration establishes defensibility across physical domains.
While Dua forecasts that full-stack vertical AI can dismantle legacy cost structures, build defensible data moats, and theoretically reduce drug discovery cycles from a decade to months, these outcomes depend on unresolved factors, including regulatory willingness to expedite approval pathways and the successful deployment of embodied AI in complex physical environments.
## Background
该内容由 YouTube 频道 Ai4 于 2025 年 8 月 28 日发布。视频简介指出，本场对谈聚焦医疗健康领域的垂直 AI 投资机遇，由 IA Seed Ventures 普通合伙人 Ed Dua 与其明星被投企业 NeuroAge Therapeutics 联合创始人兼 CEO Christin Glorioso 博士共同主讲，探讨如何借助高度聚焦的“垂直 AI 切片”切入医疗体系以实现快速增长，并重点展示了利用 AI 攻克阿尔茨海默病诊断与治疗的突破性范式。
## TL;DR
本次演讲深入拆解了以“垂直 AI 切片（Vertical AI Slices）”重构医疗健康及实体产业链的投资范式。NeuroAge Therapeutics 展示了其如何通过多模态大脑生物年龄早筛实现早期现金流造血，并借助因果 AI 与生成式分子设计攻克阿尔茨海默病“不可成药”靶点的创新闭环。Ed Dua 进一步列举了 Angstrom AI 模拟湿实验、HealthKey 秒级匹配临床患者等案例，描绘了将新药研发周期从 7 至 10 年大幅压缩的前景。最后，演讲延伸至工业金属切割与 Talis Robotics 智能导盲机器狗等案例，印证了垂直 AI 凭借细分杀手级应用颠覆传统巨头的跨行业通用性。
## Atmosphere & style
现场氛围既展现了硅谷早期风投的锐利与远见，又洋溢着动人的人文关怀。Ed Dua 表现得极具激情与感染力，以“首席拉拉队员”的幽默姿态宏观把控行业大势；Christin Glorioso 博士则以沉稳克制的科学家风范，坦诚分享了祖母患病经历与自身携带高危突变基因的创业初心，赋予了硬核技术深切的情感重量。两人的搭档在宏观商业逻辑与严谨医学洞察之间达成了绝佳平衡，传递出充满希望的变革力量。
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=4s) 垂直 AI 切片（Vertical AI Slices）是 IA Seed Ventures 的核心投资策略，通过以高度聚焦、AI 原生的单一痛点切入庞大产业链，逐步建立壁垒并向全栈扩张。
- [02:57](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=177s) 大脑生物衰老是阿尔茨海默病最核心的风险诱因（权重约 40%），若大脑生物年龄比实际生理年龄年轻 5 岁，未来患病概率可降低 6 倍。
- [02:57](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=177s) NeuroAge 采用典型的 TechBio 商业闭环：通过多模态早筛套件获取即时现金流，沉淀稀缺的高维脑衰老数据，进而反哺原创药物研发。
- [05:38](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=338s) 颠覆传统药物‘发病后治疗且伴随强副作用’的被动范式，NeuroAge 转向从根源上抗衰与提前干预的长寿医学逻辑。
- [05:38](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=338s) NeuroAge 利用孟德尔随机化与因果 AI 算法从组学数据中剖析致衰机理，针对阿尔茨海默病的高度生物异质性实现精准靶向分型。
- [05:38](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=338s) 通过与生成式 AI 药物设计机构 Klein 合作从头设计新型分子结构，精准结合以往无法成药的致病蛋白，有望将全球成药靶点数量提升 10 倍。
- [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) 当前垂直 AI 变革类似于早年本地部署向 SaaS 转型的转折点，市场正分化为老牌存量转型、全栈垂直 AI 以及纯 AI 原生三大阵营。
- [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) 传统新药 7-10 年的研发周期正被精细的垂直 AI 切片逐一重塑，长远愿景是将候选药物从发现到通过 FDA 审批压缩至数月级。
- [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) Angstrom AI 致力于将高成本湿实验全面迁移至计算机硅基模拟（In silico），Zeon Systems 则引入实验室自动化具身机器人以加速实验通量。
- [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) HealthKey 针对传统临床试验受试者招募周期漫长的痛点，通过医疗 AI 引擎直连医院病历系统，实现数分钟内的精准排他性匹配。
- [13:18](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=798s) 垂直 AI 展现出极大的应用跨度：既能以全流程 AI 深度改造传统金属切割制造厂供货头部航天企业，也能支撑通用具身智能导盲机器狗平台。
- [13:18](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=798s) Talis Robotics 以智能导盲机器狗为杀手级垂直应用，在大幅降低高昂训练成本的同时，借由实际运行沉淀极具壁垒的交互与全地形数据湖。
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=4s) 垂直 AI 切片（Vertical AI Slices）与早期投资哲学
IA Seed Ventures 普通合伙人 Ed Dua 开启了关于医疗健康垂直 AI 发展趋势的分享。Ed Dua 拥有超过 15 年的早期科技投资经验，至今已在种子轮阶段成功投出 5 家 IPO 企业并实现了多笔丰厚的退出回报。他指出，当前正处于投资史上最令人兴奋的黄金时代，而该基金的核心投资主线便是“垂直 AI 切片（Vertical AI Slices）”。这一理念的核心逻辑在于：生成式 AI 与垂直算法正在以极细分、精准的切入点全面渗透并重塑几乎所有传统行业——从工业钣金切割到视障辅助机器人，无一例外。
Ed Dua 强调，今天讨论的重心是医疗健康领域（Healthtech），并隆重引介了基金旗下的明星投资组合——NeuroAge Therapeutics 及其联合创始人兼 CEO Christin Glorioso 博士。他笑称 Christin 总是习惯性地低估自己公司的潜力和成就，因此作为投资人的重要使命之一就是担当其“首席拉拉队员”，向外界传递其颠覆性创新的真正价值。
- **投资业绩背景**：Ed Dua 从种子期投资培育出 5 家上市公司，专注深耕垂直场景中的高爆发力 AI 标的。
- **垂直切片（AI Slice）定义**：以高度聚焦、AI 原生的单一痛点切入庞大产业链，逐步建立壁垒并向全栈扩张的颠覆性商业模式。
```mermaid
flowchart TD
accTitle: 医疗健康垂直 AI 投资范式演进
accDescr: 展示 AI 切片如何通过细分场景渗透并重构医疗全产业链
subgraph Legacy["传统模式痛点"]
L1["7-10 年药物研发周期"]
L2["昂贵的湿实验与漫长临床招募"]
L3["后验式、副作用严重的重症治疗"]
end
subgraph Slice["垂直 AI 切片突破"]
S1["Angstrom AI: 计算机模拟实验 In Silico"]
S2["HealthKey: 临床试验患者秒级匹配"]
S3["NeuroAge: 大脑生物年龄早筛与因果推断"]
end
subgraph Vision["全栈颠覆与长期愿景"]
V1["研发周期缩短至数月 / 成本极低化"]
V2["不可成药靶点全面破解"]
V3["全生命周期主动抗衰与精准预防"]
end
Legacy --> Slice --> Vision
```
---
### [02:57](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=177s) NeuroAge Therapeutics：通过大脑生物年龄评估与干预预防阿尔茨海默病
NeuroAge Therapeutics 联合创始人兼 CEO Dr. Christin Glorioso 拥有麻省理工学院（MIT）神经科学博士及医学训练背景，此前曾担任斯坦福大学衍生企业的 AI 负责人，并于三年前创立了 NeuroAge。Christin 的创业初心深深植根于其个人经历：童年时期，她的祖母遭受了长达 7 年阿尔茨海默病的折磨，全家人由祖父在照料中经历了漫长而痛苦的过程，这促使她立志终生寻找根治与预防阿尔茨海默病的有效方案。在创办 NeuroAge 的过程中，她通过基因检测发现自己携带了一个 ApoE4 等位基因——这是常见的高危遗传因子（约 20% 的人群携带单拷贝），意味着她在 70 至 80 岁时患阿尔茨海默病的风险是常人的 2 到 3 倍。
然而，Christin 强调阿尔茨海默病并非不可逆转的宿命。科学研究表明，大脑生物衰老是该病最大的风险诱因，占据了约 40% 的发病风险权重。如果一个人的大脑生物年龄能够比实际生理年龄年轻 5 岁，未来患阿尔茨海默病的概率将直接降低 6 倍。作为一家典型的 TechBio（科技生物）企业，NeuroAge 将 AI 深度融入全业务流，构建了一套“早筛服务造血 + 稀缺数据飞轮 + 原创药物研发”的商业闭环：
```mermaid
flowchart LR
accTitle: NeuroAge 商业与数据飞轮
accDescr: 展示 NeuroAge 从多模态检测获取营收并反哺药物研发的完整闭环
subgraph Screening["多模态早筛与即时造血"]
A["脑部 MRI 影像"] --> D["AI 综合评估"]
B["认知功能测评"] --> D
C["专利血液生物标志物"] --> D
D --> E["出具精准大脑生物年龄"]
E --> F["个性化生活方式干预方案"]
end
subgraph DataFlywheel["稀缺数据资产沉淀"]
F --> G["多模态 / 多组学高维数据集"]
end
subgraph DrugDiscovery["创新药研发转化"]
G --> H["因果 AI 锁定致衰基因与通路"]
H --> I["联合药企开发抗衰/预防靶向药"]
end
```
- **早期现金流变现**：通过向C端/B端销售融合脑部 MRI、认知测试与专利血液标志物的“大脑年龄检测套件”获得早期收入，打破了传统生物医药企业长期零收入的困境。
- **高壁垒数据集构建**：利用检测持续收集宝贵的多模态（Multimodal）与多组学（Multi-omic）人脑衰老数据，填补全球生物医药行业在脑衰老机制领域的原始数据空白。
- **药物管线合作**：依托沉淀数据挖掘导致大脑加速衰老的核心基因与分子通路，目前已与大型跨国药企合作推进多款阿尔茨海默病预防药物的先导管线。
---
### [05:38](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=338s) 因果 AI 与生成式 AI 攻克“不可成药”靶点
Ed Dua 指出，投资界以往对早期新药研发公司普遍较为谨慎，但 NeuroAge 之所以极具吸引力，不仅在于其具备早筛带来的自我造血能力，更在于其颠覆了现有阿尔茨海默病药物“发病后治疗且伴随严重毒副作用”的被动逻辑，开创了“长寿抗衰（Longevity）与提前干预”的全新范式。
Christin 进一步阐述了 AI 在其药物研发底层逻辑中的具体运用。为了避免传统机器学习只能寻找表面关联性的缺陷，NeuroAge 运用孟德尔随机化（Mendelian Randomization）与因果 AI（Causal AI）算法，直接从复杂组学数据中剖析大脑衰老的真正因果机制。她将这一逻辑类比于肿瘤治疗：“正如并非所有癌症都相同，阿尔茨海默病也具有高度的生物异质性，必须针对个体的特定通路制定精准治疗策略。”
此外，针对生物制药行业长期面临的“不可成药（Undruggable）”瓶颈，NeuroAge 与前沿 AI 药物设计机构 Klein 达成深度战略合作。人体内约 90% 的致病蛋白由于缺乏传统活性结合位点而无法被现有药物分子靶向，而 Klein 通过底层生成式 AI（Generative AI）从头设计前所未有的新型分子结构，精准结合这些以往束手无策的致病蛋白。Christin 表示，这种技术将使全球潜在成药靶点数量实现“10 倍级的跃升（10x all of the drugs that are possible）”，为攻克神经退行性疾病提供了前所未有的武器库。
---
### [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) 缩短 10 年周期：医疗垂直 AI 切片在药物研发与临床中的应用案例
Ed Dua 将当前垂直 AI 的爆发浪潮类比为当年企业计算从本地部署（On-prem）向 SaaS 软件转型的历史时刻。彼时许多传统风投认为本地部署市场有限，SaaS 难以做大，但随着 SaaS 大幅降低软件使用成本，整个市场规模爆发了指数级增长。在垂直 AI 时代，市场将分化为三类玩家：
1. **老牌存量企业（Legacy companies）**：试图在原有体系内自我 AI 革命，但转型极度困难；
2. **全栈垂直 AI 企业（Full-stack vertical AI）**：以 AI 为底层架构从头构建，旨在全面替代传统巨头；
3. **纯 AI 原生企业（Pure AI companies）**：解决此前技术条件下根本无法解决的问题。
在传统制药领域，一款新药从候选分子确定到最终通过 FDA 审批通常需要耗费 7 到 10 年时间以及数十亿美元。Ed Dua 的投资策略是在整个漫长的生命周期中布局各个精细的“AI 切片”，逐步将各环节压缩优化，并大胆提出假想：“希望未来 10 年内，一款候选药物从确定到通过 FDA 审批仅需 3 个月——这听起来不可思议，但绝非遥不可及。”
- **Angstrom AI（计算机模拟湿实验）**：传统新药进入临床前需要耗费海量资金和时间在物理实验室进行湿实验（Wet lab experiments）。Angstrom AI 致力于将湿实验环境全面迁移至计算机硅基模拟（*In silico*），在软件中高精度模拟生物化学反应，从而大幅压缩前期筛选周期与试错成本。
- **Zeon Systems（实验室机器人自动化）**：将具身智能与 AI 机器人引入生物实验室日常操作，实现高通量实验的自动化与智能化，大幅提升化合物验证的周转速度。
- **HealthKey（临床试验患者秒级招募）**：传统临床试验（Clinical Trials）受制于极其严苛复杂的入组排他条件（如胆固醇指标、肝肾功能限制等），往往需要花费数月甚至数年时间在各大医院漫长搜寻合适受试者。HealthKey 利用专用医疗 AI 引擎直接与医院海量病历数据库对接，在几分钟内即可完成原本需要数月才能完成的精准患者人群匹配与筛选，极大化解了药物上市前最耗时的瓶颈。
---
### [13:18](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=798s) 跨行业垂直 AI 颠覆：从钣金切割到导盲机器狗平台
为了展现垂直 AI 颠覆力在广度与深度上的两极跨度，Ed Dua 在演讲尾声分享了两个医疗之外的鲜明投资案例：
```mermaid
flowchart TD
accTitle: 垂直 AI 应用跨度对比
accDescr: 对比传统制造与前沿机器人平台两端的垂直 AI 投资范例
subgraph LowTech["极简工业重构：传统制造 AI 改造"]
M1["深耕 6 年行业专家切入"] --> M2["全流程 AI 驱动运营 / 财务 / 客服 / 销售"]
M2 --> M3["引入定制机械臂高效切割铝材"]
M3 --> M4["以极致成本供货 SpaceX 等头部航天大厂"]
end
subgraph HighTech["前沿具身平台：Talis Robotics"]
R1["21岁博士天才团队底层架构"] --> R2["通用具身智能机器人操作系统平台"]
R2 --> R3["首发垂直杀手级应用：智能导盲机器狗"]
R3 --> R4["替代 $70,000 高昂导盲犬训练成本"]
R4 --> R5["沉淀视障人机交互与全地形自动驾驶数据湖"]
end
```
第一家公司代表了**对看似低门槛传统工业的极致全流程 AI 重构**。该创始人曾在一家专为 SpaceX 等航天企业切割大型铝材原料的传统金属加工厂工作了 6 年。如今，他利用全套 AI 与自动化工具从零搭建起一家全新的金属切割企业，将生产运营、财务核算、客户服务、销售拓展乃至车间机械臂全面接入 AI 调度系统，以极低的运营成本和高周转率直接向行业老牌巨头发起降维打击。
另一家公司则是 IA Seed Ventures 投资的代表科技前沿极值的具身智能平台——**Talis Robotics**。该团队由两位在 21 岁即斩获博士学位的天才科学家（现年 30 岁）创立，致力于打造通用的底层机器人技术平台。团队开发的第一个杀手级垂直应用是“智能导盲机器狗”。当前训练一只传统导盲犬的综合成本高达 70,000 美元且供给极度受限，而 Talis Robotics 不仅能够大幅降低助残成本，更在运行过程中如同特斯拉自动驾驶车队一般，持续采集复杂的现实物理世界环境数据以及视障人士与机器交互的多模态数据，从而构建出极具长期商业与算法价值的专属“数据湖（Data Lake）”。
Ed Dua 总结道，从一块铝板的智能切割，到指引盲人出行的智能机器狗，再到攻克阿尔茨海默病的生物大模型，垂直 AI 正在以无数精巧而强大的切片重塑现实世界。两人随后向全场观众致意，并正式进入问答环节。
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=4s) 垂直 AI 切片（Vertical AI Slices） — Ed Dua
> [00:04](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=4s) Christin 总是习惯性地低估自己公司的潜力和成就，因此作为投资人的重要使命之一就是担当其“首席拉拉队员”，向外界传递其颠覆性创新的真正价值 — Ed Dua
> [02:57](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=177s) 大脑生物衰老是该病最大的风险诱因，占据了约 40% 的发病风险权重。如果一个人的大脑生物年龄能够比实际生理年龄年轻 5 岁，未来患阿尔茨海默病的概率将直接降低 6 倍 — Dr. Christin Glorioso
> [05:38](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=338s) 正如并非所有癌症都相同，阿尔茨海默病也具有高度的生物异质性，必须针对个体的特定通路制定精准治疗策略 — Dr. Christin Glorioso
> [05:38](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=338s) 10x all of the drugs that are possible — Dr. Christin Glorioso
> [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) 希望未来 10 年内，一款候选药物从确定到通过 FDA 审批仅需 3 个月——这听起来不可思议，但绝非遥不可及 — Ed Dua
> [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) 老牌存量企业（Legacy companies） — Ed Dua
> [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) 全栈垂直 AI 企业（Full-stack vertical AI） — Ed Dua
> [09:19](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=559s) 纯 AI 原生企业（Pure AI companies） — Ed Dua
> [13:18](https://www.youtube.com/watch?v=AzvZYOZIz_Q&t=798s) 从一块铝板的智能切割，到指引盲人出行的智能机器狗，再到攻克阿尔茨海默病的生物大模型，垂直 AI 正在以无数精巧而强大的切片重塑现实世界 — Ed Dua
## Mentioned
IA Seed Ventures, Ed Dua, NeuroAge Therapeutics, Dr. Christin Glorioso, Angstrom AI, HealthKey, Klein, Zeon Systems, Talis Robotics, SpaceX, FDA, MIT, Stanford University, ApoE4
## Video description
Ed Dua, General Partner, IA Seed Ventures
Christin Glorioso, Co-Founder & CEO, NeuroAge Therapeutics
Abstracts
Where Will AI Impact Healthcare? Everywhere.
AI is fueling the creation of thousands of healthtech startups—some disrupting massive incumbents, others building businesses that simply weren’t possible before. In this session, we’ll dive into the concept of AI vertical slices—narrow, AI-first entry points into the healthcare system that can unlock rapid growth. We’ll also spotlight one of IA Seed Ventures’ most promising early-stage investments: NeuroAge Therapeutics, a startup using AI to develop groundbreaking Alzheimer's diagnostics and therapeutics.
[Source](https://www.youtube.com/watch?v=AzvZYOZIz_Q)
Source quotes
[0:04] Ed Dua: 垂直 AI 切片（Vertical AI Slices）
[0:04] Ed Dua: Christin 总是习惯性地低估自己公司的潜力和成就，因此作为投资人的重要使命之一就是担当其“首席拉拉队员”，向外界传递其颠覆性创新的真正价值
[2:57] Dr. Christin Glorioso: 大脑生物衰老是该病最大的风险诱因，占据了约 40% 的发病风险权重。如果一个人的大脑生物年龄能够比实际生理年龄年轻 5 岁，未来患阿尔茨海默病的概率将直接降低 6 倍
[5:38] Dr. Christin Glorioso: 正如并非所有癌症都相同，阿尔茨海默病也具有高度的生物异质性，必须针对个体的特定通路制定精准治疗策略
[5:38] Dr. Christin Glorioso: 10x all of the drugs that are possible
[9:19] Ed Dua: 希望未来 10 年内，一款候选药物从确定到通过 FDA 审批仅需 3 个月——这听起来不可思议，但绝非遥不可及
[9:19] Ed Dua: 老牌存量企业（Legacy companies）
[9:19] Ed Dua: 全栈垂直 AI 企业（Full-stack vertical AI）
[9:19] Ed Dua: 纯 AI 原生企业（Pure AI companies）
[13:18] Ed Dua: 从一块铝板的智能切割，到指引盲人出行的智能机器狗，再到攻克阿尔茨海默病的生物大模型，垂直 AI 正在以无数精巧而强大的切片重塑现实世界
Source quotes
[0:04] Ed Dua: never been more excited in terms of the opportunities
[0:04] Ed Dua: cheerleader
[5:37] Ed Dua: really awful
[5:37] Dr. Christin Glorioso: not all Alzheimer's is the same
[5:37] Dr. Christin Glorioso: 10x all of the drugs that are possible for diseases
Original transcript
[0:04] Great. Well, thanks everybody for
[0:05] coming. Um, so my name is Adue and I run
[0:09] a very early stage uh seed fund um out
[0:12] of Berkeley, California. And today we're
[0:14] going to be talking about um health care
[0:17] or health tech as it relates to vertical
[0:19] AI, what I call slices, which is
[0:22] becoming the theme of our fund, which uh
[0:25] you know, the word AI is is all around
[0:28] us every day, and it's growing and
[0:29] growing. Um and we see an is the most
[0:34] amazing time for for investing. I've
[0:37] been investing for over 15 years in
[0:39] early stage companies. Um, and I've
[0:43] never been more excited in terms of the
[0:44] opportunities. But what I mean by uh a
[0:48] vertical AI slice, what I mean is just
[0:50] about every industry is going to be
[0:51] disrupted and more so more than others,
[0:55] but you're going to see just dramatic
[0:57] change across the board. So, we're going
[0:58] to talk about that today. Um, I'm going
[1:01] to give you my perspective and what we
[1:02] do. And I brought Kristen who is one of
[1:06] our portfolio companies. Uh and she's uh
[1:09] an amazing entrepreneur. She is working
[1:13] we'll talk a little bit about what she's
[1:14] doing but uh she has an amazing um
[1:17] background uh PhD from MIT and working
[1:20] on um a very important area um in in
[1:25] Alzheimer's.
[1:26] So, we're going to relate that to our
[1:29] philosophy and what we're doing in terms
[1:31] of investing and we'll go from there.
[1:32] So, I want to give you a quick
[1:34] background in terms of who we are. So,
[1:36] I'm just going to quickly go through
[1:37] this. So, again, like I said, I've been
[1:38] investing for about 15 years and the AI
[1:43] slice is really it's going to it's going
[1:45] to hit every everything. Again, today
[1:47] we're going to talk about healthcare and
[1:49] these are all the areas and and
[1:50] anything. I don't care. It's going to be
[1:52] uh it could be a sheet metal cutting
[1:55] company or it could be a company
[1:58] building robotics for the seeing
[2:00] impaired. There's such a wide gamut in
[2:02] terms of the types of disruption we're
[2:04] going to see. So, um I've been very
[2:07] fortunate and had some very good
[2:08] successes. I've had five IPOs from seed
[2:12] stage. um a number of other very large
[2:15] uh uh returns, realized returns, and
[2:20] hopefully a bunch of these other
[2:21] companies are going to come fruition.
[2:22] So, uh that's that's what we hope. So,
[2:25] we make a lot of bets. I've made a lot
[2:26] of bets over the years um and I've been
[2:28] lucky enough to have um quite a few that
[2:30] have um that have been realized in terms
[2:33] of returns. So, my team real quick.
[2:37] [clears throat]
[2:38] Um, and so what we're going to talk
[2:39] about now is, uh, I want to turn the the
[2:42] mic over to Kristen. She's going to tell
[2:44] you a little bit about the company and,
[2:46] um, she always understates her company.
[2:49] So, I'm here to be the cheerleader
[2:51] because that's one of the big important
[2:52] roles that I have. So, Kristen, take it
[2:55] away and give us a little overview of
[2:56] what you're doing.
[2:57] >> Thanks so much. Um, uh, really happy to
[3:00] be here. So, I'm Dr. Kristen Glorioso.
[3:02] I'm a physician and neuroscientist by
[3:05] training with training from MIT. Um,
[3:08] previously I was head of AI of a
[3:10] Stanford spinouts um before starting
[3:13] NeuroAge about 3 years ago and
[3:16] everything that I've done in my career
[3:17] is really related to my own family
[3:20] history of Alzheimer's disease. So when
[3:22] I was a kid, my grandmother suffered
[3:24] with Alzheimer's for about seven years.
[3:26] My grandfather took care of her in the
[3:28] home. This had a really big impact on me
[3:30] and the rest of the family led me to
[3:32] dedicate my life to creating better
[3:34] solutions for Alzheimer's disease. And
[3:37] in the course of starting NeuroAge, I
[3:39] found out that I myself have a genetic
[3:41] risk factor for Alzheimer's disease. So
[3:43] I have a copy of the ApoE4 alil. This is
[3:46] a really common genetic risk factor.
[3:48] About 20% of people have one copy. Puts
[3:52] me at two to three times the risk of
[3:53] having Alzheimer's when I'm in my 70s or
[3:56] 80s. Um, and the good news is is that uh
[4:00] there's really something you can do
[4:02] about it. So, I've been studying brain
[4:04] aging for my whole career and uh brain
[4:07] aging is the biggest risk factor for
[4:08] Alzheimer's disease and it's about 40%
[4:11] of the risk. And if you can be 5 years
[4:14] younger biologically than you are
[4:16] chronologically, you're six times less
[4:18] likely to have Alzheimer's in the
[4:20] future. And a lot of that is influenced
[4:23] by lifestyle. Um, we're also creating
[4:26] therapeutics that can target brain aging
[4:28] itself and help people to reverse those
[4:30] changes and stay sharper and have
[4:32] Alzheimer's preventative drugs. So,
[4:35] NeuroAge um, we are a tech bio company
[4:39] which was mentioned before. We use AI in
[4:41] everything we do. Um, and we're have
[4:44] early revenue that we're generating now
[4:47] by selling a test for brain aging. It
[4:49] combines brain MRI with cognitive
[4:51] testing and blood biomarkers that are
[4:53] proprietary to the company. Um, and we
[4:56] give people a brain age so they can
[4:58] start working on their lifestyle. We
[5:00] give personalized lifestyle
[5:01] recommendations for people to slow down
[5:03] or reverse their brain aging. And then
[5:06] we're creating this super valuable data
[5:08] set um that's multimodal and multi-omic
[5:13] and a data set that doesn't exist that
[5:15] we really need for drugs uh that can
[5:19] target brain aging itself and prevent
[5:21] Alzheimer's in the future. And so we can
[5:23] use AI to figure out what genes and
[5:26] pathways people have that is causing
[5:28] their brain to be aging faster or
[5:29] slower. and we've started working on
[5:31] those therapeutics with large
[5:33] pharmaceutical partners.
[5:36] Um, that's a [clears throat] little bit
[5:37] about us,
[5:37] >> right? So, what I loved about Christian
[5:41] uh is um there's a couple things. A,
[5:44] she's building a unique company. She's
[5:47] generating revenue now. Uh, she's
[5:49] building a data set which from this
[5:52] unique testing, she already has several
[5:55] candidates for for Alzheimer's. and her
[5:58] unique approach is if you look at
[5:59] Alzheimer's today, the ones that are
[6:02] treating her, they're really awful and
[6:04] that the the the side effects are are
[6:06] are just really horrific. So Kristen's
[6:08] really focusing ahead of time longevity,
[6:12] keep ourselves healthy and creating
[6:15] therapeutics to do that. So, what I love
[6:18] about this and we don't do a lot of
[6:19] therapeutics in our company, but I love
[6:21] this company because she's generating
[6:24] revenue and she's also creating she'll
[6:26] be creating some great drugs for for um
[6:29] longevity in general uh and Alzheimer's.
[6:32] So, that's that's um that's a great
[6:34] thing. So, what we're going to do is
[6:36] talk a little bit about what I love to
[6:38] do is I like to give examples of
[6:40] companies that are in the uh in the
[6:42] space. So I I'm kind of a generalist and
[6:46] with my partners we focus in certain
[6:48] areas and what I want to talk about now
[6:50] where we focus a lot for health tech is
[6:53] on the drug discovery process um once
[6:57] drug candidates are identified. Okay. So
[7:01] um but there's a huge amount of of work
[7:03] that's done ahead of time now especially
[7:05] with AI in terms of coming up with these
[7:07] drug candidates. And one of the one of
[7:09] the philosophies that I have and one of
[7:11] our strategies is you have all these
[7:13] drug candidates are just going to be in
[7:15] the pipeline. They're going to build up
[7:16] in front. It takes seven to 10 years to
[7:19] get through the FDA process. So we've
[7:21] invested in a few companies that are
[7:23] going to help shorten that time frame,
[7:25] reduce cost so that hopefully in 10
[7:28] years when you have a drug candidate,
[7:30] you can get it through FDA in 3 months.
[7:32] Right? That sounds crazy, but I don't
[7:34] think it's too far-fetched really. So,
[7:36] we're going to talk a little bit about
[7:38] um about that. Uh this is Christian's
[7:41] team
[7:43] and um so we're going to start with the
[7:47] upfront
[7:48] um kind of using AI and drug discovery
[7:51] and I'm going to turn this over to you
[7:53] because you know this far better than I
[7:55] do.
[7:56] >> Yeah. So AI basically touches uh
[7:59] everything that we do. So at NeuroAge it
[8:02] we use causal AI to identify not just
[8:06] how fast people's brains are aging but
[8:08] what's causing it and can we create
[8:10] drugs that are personalized to that
[8:12] biology in the same way that we not all
[8:14] cancer is the same not all Alzheimer's
[8:16] is the same. So we need that um
[8:19] personalization. So that's one way we
[8:21] use um mandelian randomization causal
[8:24] inference to figure out what's actually
[8:26] causing people's brain aging. Um, and
[8:29] we're partnered with companies that are
[8:30] helping us on the on the drug side. So,
[8:32] we heard in the last panel about
[8:34] undruggable diseases. That's a real
[8:37] thing. So, there are many diseases.
[8:39] Alzheimer's is one of them where um
[8:42] there are a lot of root causes. So,
[8:45] proteins that are contributing to those
[8:47] diseases that cannot be targeted with
[8:50] drugs yet. So, about 90% of proteins
[8:53] can't be targeted with drugs. And so our
[8:56] partner Klein is helping us to design
[8:58] brand new novel drugs with generative AI
[9:01] that we've never seen before that can
[9:04] target these proteins that have not had
[9:06] drugs for them before. And you can
[9:08] imagine what that's going to do. It's
[9:10] going to have, you know, 10x all of the
[9:12] the drugs that are possible for
[9:14] diseases, which is amazing. So that's
[9:17] one example.
[9:18] So I'm going to talk a bit about some of
[9:20] the investments that we have made
[9:22] outside of Neuridge.
[9:24] Uh and again the vertical AI slices I
[9:29] liken this to the days where you had
[9:32] onrem computers in in companies and then
[9:35] we went to SAS um uh products and a lot
[9:38] of the VCs are saying well the market
[9:40] isn't really that big from on-rem so
[9:43] we're not going to invest in the SAS cuz
[9:44] it's really not going to grow. Well,
[9:46] guess what? It grew incredibly because
[9:49] prices came down and many more companies
[9:51] could in effect be able to afford these
[9:53] types of things. I think the same thing
[9:55] you'll see with a lot of these slices.
[9:57] There's two types of companies. There's
[9:58] companies, legacy companies that either
[10:01] going to have to figure out how to
[10:03] reinvent themselves with AI, which is
[10:05] going to be very difficult. um or you're
[10:07] going to have companies what we call
[10:09] vertical um full stack vert vertical AI
[10:12] where they're actually going to create
[10:14] companies that are going to displace
[10:15] these legacy companies and then you have
[10:17] the pure AI companies that never would
[10:19] have existed without AI and I'm going to
[10:22] talk about a few of those right now. So
[10:24] again the drug discovery process is
[10:26] about 7 to 10 years. So we focused on
[10:29] I'm going to give you an example of one
[10:30] company we've we invested in recently
[10:33] called Angstrom AI. So in the early
[10:35] stages of drug discovery once someone
[10:38] has selected a drug you know a larger
[10:40] company has decided to put this drug
[10:42] through um the FDA process they have to
[10:45] do a bunch of wet lab experiments and
[10:48] these are expensive um they're very
[10:50] timeconuming so what angst I is doing is
[10:53] they're looking at the various wet lab
[10:55] types experiments and they're bringing
[10:57] in silica within the computer so what
[11:01] does that do it reduces the cost it
[11:03] reduces the time frame. That's an
[11:06] example of one of these slices. They
[11:08] will then build that company and expand
[11:10] out that slice. So, we feel that there's
[11:13] going to be a lot of small niche
[11:15] companies that are going to build um an
[11:18] initial uh beach head and then expand
[11:21] out. So, it doesn't we we believe it
[11:25] doesn't have to be this gigantic company
[11:27] to start in many cases if they have a
[11:29] good uh if they have a good path using
[11:30] utilizing AI. Some of the other examples
[11:33] are um uh Xeon systems which is doing um
[11:38] robotics within the uh drug discovery
[11:40] within the lab. Uh there's a bunch
[11:42] you're using AI to basically um speed
[11:46] things up um be able to do more drugs
[11:48] quicker. Um
[11:51] what else do we have? Another another
[11:53] example is uh health key. And health key
[11:56] is right now in drug trials. If you're
[12:01] if you're trying to find people for
[12:03] clinical trials, it's extremely
[12:05] difficult. You know, you have this long
[12:07] list. Well, this person can't have a
[12:10] cholesterol level this or they can't
[12:11] have liver issues or this and that. This
[12:13] is a perfect use for AI to be able to
[12:16] would take months if not years to
[12:18] identify enough people for a trial. You
[12:20] can do it in a matter of minutes with
[12:22] AI. be able to identify, you know,
[12:25] hospital has a huge database. They can
[12:27] tap into that database with AI to then
[12:31] identify patients for the test. How much
[12:32] time we g we're going to uh save on
[12:34] that? A tremendous amount of time. So
[12:37] what does that do? That's another slice
[12:39] within the AI FDA process. So that's u
[12:45] an exciting thing. And so I wanted to
[12:47] give specific examples, but we're
[12:49] looking at the entire scope of
[12:52] businesses in this area. Um, and really
[12:56] depending on our expertise and what
[12:57] these vertical slices mean. Um, so let's
[13:02] uh let's move on. I think we're doing
[13:04] great. I think we're doing great with
[13:05] time.
[13:08] We're going to have some Q&A at the end.
[13:11] So we have eight more minutes. So let's
[13:14] see.
[13:16] I'm going to I'm just going to do real
[13:18] quickly step outside of health tech if
[13:20] that's okay with you guys. I want to do
[13:22] I want to show you the extreme examples
[13:24] between companies as it relates to what
[13:27] can what is possible. So I came across a
[13:31] company they what they're I came across
[13:34] this individual what he did he worked
[13:35] for a company what do they do they
[13:38] basically take large pieces of aluminum
[13:40] and they slice them up in pieces and
[13:42] they sell them to companies like SpaceX.
[13:44] That's their business. He has worked
[13:46] there for six years. He's now starting a
[13:49] company where he's going to build this
[13:51] company from the ground up utilizing AI
[13:54] and that means operations,
[13:57] finance. He's going to bring robotics,
[14:00] customer service, sales. So, he's going
[14:03] to build this up and he's going to be
[14:04] able to run a company much more
[14:06] efficiently at a much lower cost. Thus,
[14:09] the idea is to displace the existing
[14:11] legacy company. So this is a simple
[14:14] business and I wanted to give this
[14:15] example to to to show you as we're
[14:18] looking at things. It doesn't have to be
[14:20] some complex company. You're going to
[14:22] have disruption across the board. Uh on
[14:25] the other kind of far side, this is our
[14:28] first example, our first investment uh
[14:30] in a company called Talis Robotics. And
[14:33] with Talis Robotics, we met two folks.
[14:35] They got their PhDs at age 21. They're
[14:38] now in the ripe old age of 30. And what
[14:41] they're doing is they're being a
[14:42] building a robotics platform. Robotics
[14:45] platform to be able to have bunch of
[14:46] applications on top of it. They're
[14:48] building the first application. We've
[14:51] all seen those robotic dogs outside,
[14:53] right? They're building a a platform.
[14:57] They're building an application so a
[14:59] seeing impaired person will have a
[15:01] seeing eye dog, which now costs $70,000
[15:04] for somebody to be able to train one.
[15:07] And you're creating a data lake. Think
[15:08] about Tesla and all the autonomy
[15:10] autonomous data they pick up. The dog is
[15:13] going to be picking up the robot dogs
[15:14] and pick up all this autonomous data
[15:16] plus the increase in data lake between
[15:18] the communication between the dog and
[15:21] the seeing impaired person. Okay. So,
[15:24] they're going to create this amazing
[15:25] unique uh data lake of value and they're
[15:28] building a platform and I'm I'm giving
[15:30] you this example to show you kind of the
[15:32] the the wide range of companies between
[15:35] slicing metal
[15:37] and a robotic dog that's going to help
[15:39] seeing impaired. Okay. So, um I think we
[15:42] have a few more minutes. Maybe we'll do
[15:44] some some Q&A. Is there anything else
[15:45] you want anything you want to add?
[15:47] >> No. Happy to take questions.
[15:49] >> Okay. Let's take some questions if we
[15:50] have questions.
