Setting the file. One moment.
Skill 29 · Knowledge Base For Startups
Subchapter 29.183
references/learn/general/how-lazarus-is-solving-the-worlds-toughest-problems-with-laser-targeted-models.mdMarkdown10 KBView on GitHub
AWS Startups > Learn (opens in a new tab)
Technology can solve the biggest challenges we face today—but it takes purpose-driven innovation to make that happen. That goes beyond economic impact and demands that we focus on more than finances alone. At a time when AI is reshaping how we live and work, foundation models (FMs) built with purpose can do amazing things, from helping us track down criminal networks to improving medical care.
The startup Lazarus (opens in a new tab) is proof that a highly specialized approach can drive these powerful real-world outcomes. Whereas most organizations use broad, one-size-fits-all model building approaches that are unpredictable and can’t scale, Lazarus’s method is different. It targets specific issues before creating FMs to tackle the problem head on. To date, Lazarus’s achievements include helping to prevent domestic terrorist attacks and processing approximately one-third of all US medical records for healthcare claims.
As Alex Panait, Chief Strategy Officer at Lazarus, says: “Sometimes technology is not always the answer. The answer is how do you apply it to problems?” He continues: “Instead of trying to build a model that understands everything about everything, we decided to build one that’s really good at doing one thing.”
Working with AWS, Lazarus is supercharging the human impact of its models for leading global organizations. “AWS goes beyond just giving us access to the tools, the GPUs, the infrastructure. It really brings the customer’s problems closer to us and helps us with the right strategies to tackle them,” says Panait.
Inspired to ignite positive change, the business’s journey began with the challenge of solving diagnostic problems with melanoma—one of the deadliest cancers. Being able to distinguish the word ‘urgent’ on authorization requests is critical to providing timely patient care, from medications to organ transplants, and elective surgeries. But as Panait explains, this data was trapped in biopsy reports because “physician handwriting is notoriously hard to figure out,” second only to “Egyptian hieroglyphs.” Lazarus built a specialized FM to unlock this unstructured data and understand nuanced documents with potentially life-saving impacts.
Since then, its product strategy has been driven by gaining a native understanding of information across multiple modalities. When information comes in many shapes and forms, making sense of it relies on contextualizing and interpreting data holistically. In contrast, conventional FMs are often built using a conversational approach, which limits their ability to address a specific problem. As Panait explains, “The world around us is more than text. It has images and video. It has sound. It has geospatial data, it has signal data. All of those contribute to understanding an entity or an event.”
What makes Lazarus so unique is that its AI lab designs, develops, and deploys its models from scratch. This allows the business to close feedback loops and understand the externalities of what it’s building. Panait says that its holistic approach and pace of development can make the difference between good and bad outcomes: “In life-or-death situations, you don’t have time to fine-tune. You don’t have time to retrain. What you need is to respond to reality as fast as possible.”
Lazarus roots its FMs in real problems so that it can deliver highly accurate and timely outcomes. Both subject matter experts within its business and AWS industry specialists play a crucial role in understanding these problems intimately. Commenting on the collaboration with AWS, Panait says, “One of the things that I really liked from the beginning was the genuine curiosity that the AWS team has shown. They’ve been willing to really understand the fundamentals.” With support from AWS, the startup can access vertical insights and better understand customer friction points.
By joining forces, Lazarus has been able to amplify its impact across a breadth of verticals, including insurance, manufacturing, and healthcare, as well as the public sector. It now works with 15 of the top 20 global insurers and over 40 Fortune 500 companies. “Roughly 65 percent of insurers are on AWS,” says Panait. “That industry knowledge really helps to declutter the landscape. It’s also a powerful resource that, when combined with solutions like Amazon SageMaker, can drive innovation and accelerate customer outcomes.”
For example, insurers contend with decades of aggregated data, meaning processing claims efficiently has been an ongoing challenge. Having unpicked the problem, Lazarus developed and deployed a sophisticated FM using Amazon SageMaker that mimics the reasoning of a trained analyst. By doing so, a provider was able to streamline disability claims processing from three weeks to under 30 minutes—helping people access essential funds when they need it most while helping insurers avoid late payment penalties.
The technical challenges of solving the world’s toughest problems don’t stop at creating the right model to meet specific needs. Lazarus’s models are “GPU hungry,” says Panait. “Some of our largest deployments require about four nodes of h100. Now, where do you find that? It’s just not lying around everywhere,” he adds. With world-leading organizations relying on the business to quickly create proof-of-concepts and scale solutions, on-demand compute capacity accessed through Amazon EC2 (opens in a new tab) is critical to success.
Panait says that “AWS has the depth, breadth, and geographical dispersion of these assets that allows us to serve customers from the United States all the way down to Taiwan, Australia, and Japan.” Beyond taking advantage of scalable infrastructure on a global level, Lazarus is seeking to further extend its reach through co-marketing and co-selling support. Panait describes listing on AWS Marketplace (opens in a new tab) as “a conduit for shorter time-to-value”, helping customers to access solutions rapidly while ensuring “non-technology related challenges like governance actually get solved much faster.”
As well as bringing even more practicality to customers, a key focus of the partnership is developing Lazarus’s specialized go-to-market strategy and fostering new connections. This includes everything from creating account-based strategies and tailored collateral to gaining exposure to new customers by presenting at AWS events. “We need the trust that AWS is brokering between the tech and the actual buyers of the tech. AWS is instrumental in building this market motion quickly by connecting the right folks from the business side, the solutions side, and the technology side,” says Panait.
Working in such highly regulated and confidential areas means that security, privacy, and explainability are essential for gaining and retaining customers. Whether it’s meeting strict government requirements in the defence space or navigating evolving regulations for commercial companies, Lazarus is able to maintain trust throughout model development and deployment. “AWS has been a tremendous partner in helping us deploy these technologies in air-gapped environments and really secure, safe, and trusted setups that are consequential and highly relevant to our partners,” says Panait.
The business has also designed its service so that it can quickly but safely create value from internal and external data by pretraining models and ensuring customer data is never retained. With no need for data labelling, training, or fine-tuning, customers can rapidly deploy models in their own secure AWS environments. FMs are “trained to understand aspects of the world so customers don’t have to,” says Panait. The models have three layers of explainability which mitigates hallucinations, allowing customers to confidently source information, rely on the accuracy of outputs, and substantiate subsequent actions.
Looking ahead, Panait sees the AWS partnership as a core driver for building momentum with its custom-built and outcome-oriented models: “AWS is and will be a critical part of our mission.” He adds, “It has already shown a lot of value to us in helping us build this motion that’s rooted in one North Star of solving customer problems.” With no shortage of existing and emerging challenges, Lazarus is doubling down on the world’s toughest problems one FM at a time. One powerful application Panait sees in the future is the potential for contextualized models to “fight deepfakes,” because it’s “impossible to alter context ad infinitum.”
Noting that “it takes an army to bring this into all possible applicable problem domains,” Panait says the business has “big plans with AWS.” While it started working in insurance, the startup has proven that it’s possible to rapidly expand its services not only into other areas of financial services, but healthcare, legal, semiconductor, shipping, and manufacturing spaces too. With seemingly unstoppable opportunities for growth, Lazarus will continue to act on its belief that it isn’t features or technology alone that wins trust—it’s a track record of driving real outcomes.