MemWise Pendant

Dementia is a devastating neurological disorder that affects over 10% of Americans above the age of 65, according to a 2022 study in JAMA Neurology. Besides the emotional and intellectual toll of cognitive decline, dementia also has a significant financial cost. On average, the health care cost for dementia patients is nearly $60,000 per year for their last five years of life. 

One significant problem is that dementia is often detected late, and patients miss out on the crucial timeframe of early disease progression when treatments are most effectively administered to minimize the disease effects. By providing a tool to monitor dementia onset in elderly patients, we help families maximize the time spent and memories made with their loved ones. 

Our solution securely transcribes and stores voice data for our AI to detect early dementia symptoms. Traditionally, early detection of dementia can be difficult to recognize without frequent neurological and cognitive exams. Our vision is to monitor patient neurological health through an unobtrusive, fashionable pendant worn around the neck. In addition to early detection, we believe that for the especially elderly, there are additional benefits to this rich audio data, including:

- Memoirs generated through LLMs to share with their loved ones
- Preventative evidence against elderly abuse at senior care facilities

We are particularly excited about the potential of MemWise to serve multiple, complementary roles in the lives of patients and their families.

Our solution serves the elderly population, particularly those at risk of developing Alzheimer’s and other dementia-related diseases. By providing an early notification of disease progression, patients can enroll earlier in treatment programs that minimize disease effects in the early and mid stages of dementia, improving their quality of life over a period of years. Furthermore, it provides awareness to families of how their loved ones will change in the future, allowing them the opportunity to enjoy their healthy time together. 

Together, Jason and Daniel bring a unique set of research and business skills to deliver on a solution to this problem of early-onset dementia. 

Jason brings 5+ years of applied AI/ML experience, having received an MS from Stanford where he conducted research on adapting LLM models for novel tasks, and received a Best Project Award for improving natural language processing (NLP) models. He has previous experience working on data science and applied AI problems within the healthcare space, and is excited to transfer this deep technical expertise to building and improving upon AI models for MemWise.

Daniel brings 5+ years as a private equity investor and 2+ years as a technology entrepreneur, most recently in voice and AI. He has built and demoed LLM-based SaaS applications for vertical AI use cases in the healthcare space, including for insurance authorization calls and medical appointment verification calls. He has built relationships with leading transcription, voice synthesis, and telephony software providers, and has worked closely with lawyers on how to compliantly handle sensitive patient information. He is excited to bring his experience in healthcare, entrepreneurship, and investing to help develop MemWise.

Moreover, Daniel has felt this problem firsthand through his father’s church, which ministers to elderly Korean-American immigrants. Quite a few of the church’s parishioners suffer from dementia, and it is readily evident that the economic and emotional damages that stem from dementia affect not only patient families, but also their broader communities. To play a small part in solving this problem would be incredibly meaningful. 

As a team, Jason and Daniel are confident they can tackle this problem. They bring the appropriate technical and research capabilities, as well as the entrepreneurial experience and personal empathy for the end user to successfully bring this product to market. 

Typical screens for dementia range from neurological and cognitive screens to invasive cerebrospinal fluid extractions or time-consuming and costly MRI and CT scans. These diagnoses require numerous in-person visits with medical specialists. 

Our solution is innovative in that it replaces these costly trips to the doctor with a non-invasive monitoring system that periodically records audio spoken by the patient, and uses a proprietary AI for natural language processing (NLP) system to analyze the speech patterns to determine whether the patient is exhibiting early symptoms of dementia. Recent advances in machine learning and AI have demonstrated that it is possible to reliably detect dementia symptoms from purely audio signal (including doi:10.1186/s13195-021-00888-3 and doi:10.3390/s22239311). We leverage these findings to develop our own model for analyzing speech data to detect abnormalities that signal early dementia symptoms. 

Our solution enables countries both developing and developed to better detect and treat a prevalent disease in the elderly, a typically vulnerable segment of the population. Especially in developing countries, expensive procedures and equipment for definitive detection of Alzheimer’s and other dementia-related diseases may prohibit average patients from receiving the medical care they need. Our technology only requires access to a smart phone and has a relatively small digital footprint, similar to the data necessary to transmit an internet-based phone call. Specifically, we believe MemWise matches the UN Sustainable Development Goal Target 3D, of strengthening the capacity of all countries, in particular developing countries, for early warning, risk reduction and management of national and global health risks.

Our system consists of four main components: a bluetooth enabled microphone, a smartphone app, a cloud-based backend to process data, and an AI model to determine whether customer speech exhibits symptoms of dementia.

The audio recorded from the microphone will be saved to the smartphone via the associated app. The audio will be periodically uploaded to the cloud backend, where the AI model will process the audio and generate a prediction of dementia symptom likelihood. The result of this prediction will be downloaded by the app on the user’s smartphone. To enable connectivity between family members and care providers, users can connect their profiles in the app, and approved connections can view the health status of the monitored customers.

Our deployed AI model will be based off of a state-of-the-art pretrained Transformer LLM (Whisper, LLaMA, or GPT-4).

Our training will go through two stages. First, we will finetune this LLM on a corpus of audio speech and transcripts from existing databases of speech by patients with dementia (e.g. TalkBank). This provides a baseline to analyze any given segment of speech as exhibiting signs of dementia.

Secondly, we will work with clinicians to collect anonymized training data of speech over time by patients progressing through dementia symptoms to improve the model. Furthermore, we will use the data collected from our customers to finetune the models for each specific customer, so that we may detect variations in these speech patterns over time. We recognize that there are privacy and safety concerns about using medical data, and we are committed to following health and ethical guidelines (e.g. HIPAA) for protecting patient data.

We recognize the importance of, and are committed to, ensuring the ethical and responsible use of AI for addressing our users’ needs. We believe there are three main components to ethical AI use:

• Data Privacy: All user interactions will be encrypted and stored securely to safeguard privacy. We will communicate to lawyers about maintaining compliance with all data protection laws and regulations, including HIPAA and CCPA. 

• User Consent: Users will be informed about the data collection policies before interacting with MemWise. Ensuring full oversight over how their data will be used is critical for building trust with our end users. 

• Human Oversight: AI tools are meant to assist, not replace, humans. As powerful as MemWise’s AI models will be, we believe in always including a human in the loop to analyze and interpret the data when needed. 

Our early product goals relate to product prototyping and validation, and our broad impact goals revolve around the widespread usage of MemWise. We believe long-term customer usage is the single best indicator of whether MemWise is materially benefitting the lives of the people we hope most to impact. 

Within the next year (2024), our goal is to have 1,000 daily active users of MemWise. Through the Cure Residency, we seek to partner with advisors not only for prototyping our hardware but also for marketing a consumer product to get our product to market within the next 6 months. 

Over the next five years (2024-2028), our goal is to have 1M+ daily active users of MemWise. It is a tall order, but one that we hope to achieve through medical and cultural partnerships. On the medical front, we seek to partner with 5+ leading medical institutions to help spread adoption of MemWise while also validating the diagnostic efficacy of MemWise compared to known methods. 

On the cultural front, we seek to partner with 5+ leading athletes, celebrities, and/or brands who have built credibility with our user base to champion the benefits of preventative medical care. We believe that, in order to influence consumer behavior to adopt a new wearable technology, MemWise must craft a compelling brand story around its values and benefits. 

2 people, full-time.

3 months. 

We believe the best ideas come from diversity, and the best products come with the customer in mind. To that end, we are committed to incorporating DEI principles into our work from the foundation. We would like to share a few of our principles: 

1. Team Diversity

We believe the best ideas come from people with diverse backgrounds. Different communities have different needs, so we seek to recruit employees and peers who can help teach us about the communities and social structures they come from. We will emphasize our focus on working with people from diverse backgrounds but all rooted in a common mission: serving the needs of the elderly. 

2. Inclusive Design

Our vision for the pendent is that it be user-friendly, accessible, and loveable. We have to earn the trust of our users. To that end, we hope to bring on advisors and collaborators from the elderly community across gender, races, and abilities to weigh in on design decisions that properly balance form and function. 

3. Accessibility

We understand that, despite our best efforts to keep MemWise affordable, our product may find itself in the hands of certain segments of the elderly community with greater economic means. We plan to engage in partnerships with values-aligned organizations to offer discounted pricing, rental services, or any number of other initiatives to get MemWise in the hands of those who need it most. 

The two founders of the team will be:

• Daniel Kim (CEO)

• Jason Qin (CTO)

Our first goal is to bring on a hardware advisor / consultant who can help navigate the early R&D processes of our product. 

We hope to build the organization as a user-first company. We believe we have to earn our users’ trust with the quality of our product, and to that end, will build out an advisory board of at-risk elderly and their family members. This board will continually offer its feedback on product development.

We do not anticipate being profitable for the first phase (4-6 months) of R&D as we prototype the hardware and software and build out our proprietary LLM. Our aim is for widespread adoption of our hardware, so we plan to charge $49 for it. Longer-term financial sustainability will come in the form of annual subscription payments of $99. 

In the US alone there are over 55 million adults over age 65, of which around 6 million suffer from dementia-related diseases. To reach this user base, we have various go-to-market strategies, which include: 

1. Reaching the children of elderly citizens, who may be more open to adopting emergent technologies for preventative medical care

2. Reaching the employers of elderly citizens, who have larger budgets for bulk purchases and potential subsidies for our product

3. Reaching elderly citizens directly through targeted advertising, free educational seminars, community events, etc. 

• Human Capital: $50K salary per founder, $100K total

• Hardware development: $5000

• Software infrastructure development and maintenance: $2500

• Data acquisition: $5000

• LLM training: $2500

• Total cost: $115000 annually

We are seeking $100K in funding to help cover R&D costs for hardware, software, and LLM development, as well as a living stipend for the two founders. We are hoping to release a product to market within 6 months of receiving the grant, and with that initial traction, plan to raise a seed round of equity financing from mission-aligned venture capitalists and various stakeholders. 

The Cure Residency program offers a unique platform to engage with experts across diverse fields of both healthcare and AI. We seek talented peers and mentors to help provide direction and advice for the company. Specifically, we seek: 

1. Advisory engagements with mechanical and electrical engineers who can advise on prototyping our hardware product

2. Access to clinical partners and lawyers who can guide our approach to acquiring and handling user data

3. Connections with business leaders who can advise us on challenges and opportunities for scaling the company, particularly in the B2C space

But most of all, we are excited about the opportunity to learn from and develop deep relationships with other Residents who are similarly committed to solving America’s most challenging health care problems with AI. For any idea to take root and have meaningful impact, it requires the blood, sweat, and tears of many individuals; it requires a community. We can’t wait to find such a community of supportive, intelligent, and entrepreneurial Residents as part of the Cure Residency.