<div>
  <ul>
    <li>
      <p>
        Chronic care management company Found has launched Found Assistant, a GenAI guide within the Found app. The feature had a soft launch in September 2023, where almost 33% of beta users had multiple conversations with Found Assistant.
      </p>
    </li>
    <li>
      <p>
        Found Assistant is available 24/7 to provide relevant educational information, personalized fitness routines, new recipes, lifestyle tips, and curated guidance based on member prompts, health histories, and personal preferences. Members can use an open-type field to prompt the Found Assistant with specific questions or use preset buttons to populate prompts for recipes, workout routines, meal plans, and habit-building.
      </p>
    </li>
  </ul>
  <p>
  </p>
</div>


<div>
  <p>
    <a href="https://www.businesswire.com/news/home/20231025201328/en/Found-Taps-Generative-AI-to-Power-Instant-Personalized-Weight-Guidance-with-Launch-of-Found-Assistant%E2%84%A2">
      Press release
    </a>
  </p>
  <p>
    <a href="https://hitconsultant.net/2023/10/26/found-launches-generative-ai-guide-for-personalized-weight-guidance/">
      Hit Consultant
    </a>
  </p>
  <p>
  </p>
</div>


Preventive Healthcare

DISRUPTOR_PRESENT

PROMINENT_CATEGORY

The Chronic care management company called Found has introduced Found Assistant, a GenAI guide within their app. In September 2023, the feature was soft-launched and around 33% of beta users engaged in multiple conversations with Found Assistant. Found Assistant is accessible 24/7 and offers relevant educational information, personalized fitness routines, new recipes, lifestyle tips, and curated guidance. It provides customized support based on member prompts, health histories, and personal preferences. Members can use an open-type field to ask specific questions or select preset buttons for prompts related to recipes, workout routines, meal plans, and habit-building.

Found launches GenAI tool for its weight management app

Empowering Tomorrow's Innovations with Generative AI Infrastructure

Generative AI Infrastructure

INCUMBENTS_PRESENT

<div>
 <p>Amazon Web Services (AWS) has introduced an innovative tool called Automated Reasoning checks as part of its Bedrock model hosting service. This tool has been purposed to mitigate AI hallucinations and uses a validation system that checks model responses against information supplied by customers.</p>
 <p>These Automated Reasoning checks function by comparing AI responses with ground truth data from customers. These checks produce rules which can be honed and applied to various models. While generating responses, they carry out cross-verification and at the same time display correct responses in conjunction with possible falsehoods for contrast.</p>
 <p>AWS declares that it is the first among major cloud providers to offer a safeguard effect against AI hallucinations. Moreover, it empowers clients to establish as well as personalize safety, privacy and truthfulness within an all-in-one solution. PricewaterhouseCoopers (PwC) is already incorporating this tool in producing AI assistants for its clientele. However, AWS hasn't shared any data that demonstrate the tool's reliability.</p> 
</div> <p>Analyst <b>QuickTake:<br>
 </b> The update builds on AWS' existing hallucination detection capabilities, initially having Guardrails for Amazon Bedrock in <a href="https:<br>
//sp-edge.com/updates/24413"> November 2023 </a> . In <a href="https:<br>
//sp-edge.com/updates/31381"> July 2024 </a> , the company announced contextual grounding checks (to detect hallucinations in model responses based on a reference source and user query) and ApplyGuardrail API, which allows customers to evaluate input prompts and model responses against configured safeguards for any foundation model.</p>

AWS launches Automated Reasoning checks to combat AI hallucinations

Managing employee health-related benefits

Health Benefits Platforms

<div>
 
 <p>Soda Health, a benefits management provider, has raised USD 50 million in a Series B funding round. This was led by General Catalyst, along with contributions from Bruce Broussard, Lightspeed Venture Partners, Define Ventures, Qiming Venture Partners US, and SVB Capital, increasing the company's total raised funds to USD 81 million.</p>
 <p>The funding will be utilized to scale operations, introduce new data products, and develop capacities to new compliance management systems. This includes CMS-compliant benefit categories and gap closure strategies. The ambition is to establish risk-based structures with Medicare Advantage and Medicaid plans and widen CMS-compliant benefit offerings.</p>
 <p>As of December 2024, Soda Health's Smart Benefits platform operates in over 50,000 retail locations in the US, including prominent chains such as Kroger, Albertsons, CVS, and Hy-Vee. In the same year, over 33 million seniors relied on Medicare Advantage, accounting for USD 131 billion in supplemental benefits and reward dollars.</p>
 
</div>

Soda Health raises USD 50 million in oversubscribed Series B funding to scale operations

<div>
 
 <p>StretchDollar, a health benefits platform, has garnered USD 6 million in a seed funding round. Fika Ventures and Oscar Health were the lead investors, with Precursor Ventures and Springbank also participating. This amplifies the total funds secured by StretchDollar to USD 7.6 million.</p>
 <p>The capital will be utilized to augment the company's self-service platform and refine the benefits experience for small businesses.</p>
 <p>The platform of StretchDollar capitalizes on the Individual Coverage Health Reimbursement Arrangements (ICHRAs) rule instated by the Internal Revenue Service (IRS). This rule enables employers to allocate pre-tax dollars towards employee healthcare expenses. As of December 2024, the company offers its platform to small businesses without any charges.</p>
 
</div>

StretchDollar raises USD 6 million in seed funding to expand platform

A concentration of wealth and a willingness to adopt age-in-place technologies present new opportunities.

Age Tech

<div>
 
 <p>Digital care coordination company, ThoroughCare, has successfully integrated its care management software into PointClickCare's platform. This strategic integration merges ThoroughCare's care coordination offerings with PointClickCare's cloud-based elder care software. As a result, long-term and post-acute care providers can now access comprehensive patient data including demographics, vital signs, chronic conditions, and medication records, all from a unified platform. </p>
 <p> The key features of this integration include a single sign-on capability and the creation of specific care plans facilitated by ThoroughCare's assessment tools. Furthermore, these care plan reports are effortlessly forwarded to patient charts on PointClickCare’s platform, optimizing the overall care process. This partnership aims to streamline care management, making it more efficient and effective for all parties involved.</p>
 
</div>

ThoroughCare integrates care management software with PointClickCare for coordinated healthcare delivery

Empowering sustainable food innovation through food engineering tech

Alternative Ingredients

ANALYST_QUICK_TAKE

<div>
 <p>Fraunhofer IME, metaX Institut für Diätetik GmbH, and Candidum GmbH have collaborated on the NovelSweets project to develop a new protein-based sweetener, X3. This initiative comes as part of the German Federal Ministry of Food and Agriculture's efforts to decrease sugar content in ready-to-eat foods. </p>
 <p>Brazzein, a sweet protein in the African plant, Pentadiplandra brazzeana, is used to create X3, which is reportedly 10,000x sweeter than regular table sugar. The production process embeds sweet-tasting protein genes into yeast cells, multiplies them in a bioreactor, and then refines and dehydrates the product.</p>
 <p>X3 is a calorie-free sweetener, posing no risk of cavities or raised blood sugar levels. MetaX is at the moment developing the inaugural range of X3 products, with a cocoa-based beverage powder being the first. Prior to seeking regulatory approval, further optimization of the manufacturing process is necessary. </p>
</div> <p><b>Analyst QuickTake:<br>
 </b> Given the product's high potency, this development marks a major milestone for the food tech industry. During the year, <a href="https:<br>
//sp-edge.com/updates/29597"> Sweegen </a> and <a href="https:<br>
//sp-edge.com/updates/36552"> MycoTechnology </a> launched brazzein and honey truffle sweet protein products, although these showcased much less sweetness than X3 (around 500x–2,500x).</p>

NovelSweets project develops protein-based sweetener 'X3'

Foundation Models

<div>
 <p>Amazon Web Services (AWS) is collaborating with Anthropic in developing Project Rainier, an immense AI supercomputer, equipped with Trainium 2 chips from AWS.</p>
 <p>This supercomputer is anticipated to become the biggest reported AI machine worldwide, with its size being 5 times larger than the current cluster owned by Anthropic.</p>
 <p>An enormous number of AWS's AI training chips, alongside Trn2 UltraServer clusters, will be used in this project. This provides a financial advantage as it offers cost savings between 30% and 40% compared to the GPU-based solutions from NVIDIA.</p>
</div> <p><b>Analyst QuickTake:<br>
 </b> Amazon has invested in Anthropic, most recently including a USD 4 billion <a href="https:<br>
//sp-edge.com/updates/36441"> round </a> to become the latter's primary cloud and training partner. The funding will enable the company to manage its cash burn, reportedly <a href="https:<br>
//www.theinformation.com/briefings/anthropic-projected-to-burn-more-than-2-7-billion-in-cash-this-year"> projected </a> to be USD 2.7 billion in 2024. Despite this financial backing, Anthropic has faced challenges. In <a href="https:<br>
//techcrunch.com/2024/11/04/anthropic-hikes-the-price-of-its-haiku-model/"> November 2024 </a> , the company unexpectedly increased the price of one of its models and <a href="https:<br>
//the-decoder.com/anthropic-delays-next-gen-ai-model-opus-3-5-with-no-new-timeline/"> delayed </a> the launch of its anticipated next-generation model, Claude 3.5 Opus.</p>

Amazon partners with Anthropic to build world's largest AI supercomputer 'Project Rainier'

<div>
 <p>Amazon Web Services (AWS) has teamed up with Anthropic for Project Rainier, the development of a high-level AI supercomputer that leverages AWS' Trainium 2 chips. Representing a significant expansion from Anthropic's existing setup, the supercomputer is projected to be the world's largest reported AI machine.</p>
 <p>The project will greatly benefit from the hundreds of thousands AWS' latest AI training chips and Trn2 UltraServer clusters. The use of these resources is estimated to provide significant savings in the range of 30% – 40% compared to alternatives based on NVIDIA's GPU.</p>
</div> <p><b>Analyst QuickTake:<br>
 </b> Amazon has invested in Anthropic, most recently including a USD 4 billion <a href="https:<br>
//sp-edge.com/updates/36441"> round </a> to become the latter's primary cloud and training partner. The funding will enable the company to manage its cash burn, reportedly <a href="https:<br>
//www.theinformation.com/briefings/anthropic-projected-to-burn-more-than-2-7-billion-in-cash-this-year"> projected </a> to be USD 2.7 billion in 2024. Despite this financial backing, Anthropic has faced challenges. In <a href="https:<br>
//techcrunch.com/2024/11/04/anthropic-hikes-the-price-of-its-haiku-model/"> November 2024 </a> , the company unexpectedly increased the price of one of its models and <a href="https:<br>
//the-decoder.com/anthropic-delays-next-gen-ai-model-opus-3-5-with-no-new-timeline/"> delayed </a> the launch of its anticipated next-generation model, Claude 3.5 Opus.</p>

<div>
 <p>Amazon Web Services (AWS) has introduced multi-agent collaboration capabilities for Amazon Bedrock, now in preview. This initiative allows businesses to construct and control AI agents that can cooperate to undertake complex tasks, made available through the Amazon Bedrock platform.</p>
 <p>Thanks to this new feature, businesses can create specialized agents for various tasks, along with a supervisor agent to oversee them. It offers easy setup minus the intricate coding requirement, the ability to incorporate existing agents, and efficient inter-agent communication. It also supports optimized collaboration modes with supervisor and routing options, and features an integrated trace and debug console for dissecting multi-agent interactions.</p>
 <p>The multi-agent collaboration brings an enhancement in task achievement rates and precision compared to single-agent systems. AWS references Moody's adoption, where a group of specialized agents evaluated different dimensions like macroeconomic patterns and company risks, resulting in a finer risk assessment.</p>
</div>

AWS launches multi-agent collaboration capabilities for Amazon Bedrock

The toolkit to build machine learning applications at scale

Machine Learning Infrastructure

<div>
 <p>During the re:Invent 2024 conference, Amazon's AWS made announcements regarding the availability of Trainium2 (T2) chips, specifically for training large language models. These chips, located in each EC2 instance, deliver a powerful compute performance of up to 20.8 petaflops. AWS also unveiled Trainium3, a chip built on a three-nanometer process, which is designed to offer 4 times the performance of Trainium2. This new chip is planned for release in late 2025.</p>
 <p>EC2 Trn2 UltraServers have the capacity to deploy Trainium2 chips, featuring an impressive 64 interconnected chips able to reach up to 83.2 peak petaflops of compute. AWS uses the NeuronLink interconnect in these UltraServers to link the Trainium chips. Together with Anthropic, a giant chip cluster is being built by AWS incorporating hundreds of thousands of these chips.</p>
 <p>AWS claims that in comparison to current chips, Trainium2 offers a price performance that's 30% to 40% better and has the ability to produce token-generation throughput that's three times superior to other cloud providers.</p>
</div> <p><b>Analyst QuickTake:<br>
</b> While the AWS chips outperform NVIDIA's current-gen offerings, their performance falls short of the company's next-gen Blackwell chips (offering up to 720 petaflops of FP8 performance in a rack with 72 Blackwell GPUs). However, the chip allows Amazon to capitalize on the growing demand for GenAI infrastructure, with Blackwell's general availability <a href="https:<br>
//www.datacenterdynamics.com/en/news/nvidia-delays-blackwell-gpus-impacting-hyperscaler-data-center-plans-report/"> delayed </a> until 2025.</p>

AWS announces Trainium2 chip general availability and upcoming Trainium3 chip at re:Invent 2024 conference

<div>
 <p>At the re:Invent 2024 conference, Amazon's AWS announced the availability of Trainium2 (T2) chips, designed to train and deploy large language models. One EC2 instance with 16 T2 chips can generate up to 20.8 petaflops of compute performance. Furthermore, the release of Trainium3, expected in late 2025, promises to enhance performance fourfold thanks to a three-nanometer process design.</p>
 <p>The EC2 Trn2 UltraServers can deploy Trainium2 chips. Each server houses 64 interconnected chips and can achieve up to 83.2 peak petaflops of compute. This is made possible by an interconnect known as NeuronLink, connecting all the Trainium chips. Furthermore, AWS is constructing a large cluster with Anthropics, incorporating hundreds of thousands of these chips.</p>
 <p>According to the company, the Trainium2 chips offer improved price performance, with a 30% to 40% increase compared to existing chips. Additionally, these chips can generate tokens at a rate three times faster than competing cloud providers.</p>
</div> <p><b>Analyst QuickTake:<br>
</b> While the AWS chips outperform NVIDIA's current-gen offerings, their performance falls short of the company's next-gen Blackwell chips (offering up to 720 petaflops of FP8 performance in a rack with 72 Blackwell GPUs). However, the chip allows Amazon to capitalize on the growing demand for GenAI infrastructure, with Blackwell's general availability <a href="https:<br>
//www.datacenterdynamics.com/en/news/nvidia-delays-blackwell-gpus-impacting-hyperscaler-data-center-plans-report/"> delayed </a> until 2025.</p>

All Updates

Found launches GenAI tool for its weight management app

AWS launches Automated Reasoning checks to combat AI hallucinations

Soda Health raises USD 50 million in oversubscribed Series B funding to scale operations

StretchDollar raises USD 6 million in seed funding to expand platform

ThoroughCare integrates care management software with PointClickCare for coordinated healthcare delivery

NovelSweets project develops protein-based sweetener 'X3'

Amazon partners with Anthropic to build world's largest AI supercomputer 'Project Rainier'

Amazon partners with Anthropic to build world's largest AI supercomputer 'Project Rainier'

AWS launches multi-agent collaboration capabilities for Amazon Bedrock

AWS announces Trainium2 chip general availability and upcoming Trainium3 chip at re:Invent 2024 conference

AWS announces Trainium2 chip general availability and upcoming Trainium3 chip at re:Invent 2024 conference

Preventive Healthcare

Found launches GenAI tool for its weight management app

Contact us