Our Research Philosophy

The human mind is the engine of new ideas, new questions, new directions. But the scientific method is what separates insight from illusion. It introduces discipline. It demands validation. It separates the wheat from the chaff—and through iteration, prunes hypotheses until truth emerges. Research is not a department at Sword. It is core to what we are and how we build trust.

Our research percolates through all stages of product development. It answers four fundamental questions: Do our solutions work as intended? How do they compare to best practices? What happens when deployed at scale? Can they make healthcare sustainable? These questions guide everything we do.

Frontier AI Research

Today, AI creates value in two dominant ways: information systems that answer questions, and coding agents that build software.

In healthcare, however, the goal is not to answer a question or complete a task. It is to provide continuous care that improves health over weeks and months. Success emerges from sustained engagement across dozens of interactions. This is harder than anything current AI systems are built to solve.

Solving it requires meeting the patient where they are — understanding their history, perceiving their present state, and guiding their recovery. We are pushing the boundaries in AI research to make this possible:

Memory. Who is this patient? What have we learned across sessions — what worked, what didn't? We are building operational memory systems that drive action, not just store facts.

Perception. How is the patient doing right now? Current models struggle with raw, noisy, continuous biological data. We are building models to fuse video, audio, and wearable signals into a unified patient state — giving AI the sensory awareness to understand how a patient is actually doing.

Planning. Given who they are and where they are, how do we guide recovery? General LLMs fail at long-term strategy. Healthcare demands AI that maps a conversation today to a clinical outcome weeks from now. This requires advances in multi-turn reinforcement learning — optimizing for measurable health improvements over the full arc of treatment.

These challenges are interconnected. And solving them in one area of care unlocks them for others. Our frontier research spans benchmarks, training methods, and safety systems. All in service of AI that can actually provide care.

Controlled Trials

But getting AI right is far from being the full story. We need to prove it works. Credibly.

That’s why we do not compare against waiting lists. Or against watchful waiting. Or against sub-par care. That is not what we are trying to solve for.

Our controlled trials pit our solutions against active control groups representing the best practices in each field. This is a much higher bar. But also the only way to know if we are on par with or actually better than best-in-class care. Rigorous experimental design isolates the impact of the intervention, minimizes confounding factors, and represents the gold standard of clinical validation. We hold ourselves to it.

Real World Evidence

The real world is varied and unpredictable. That’s why controlled trials are not enough.

Controlled trials prove efficacy. Real world evidence proves effectiveness.

To prove effectiveness, we need to validate our solutions outside controlled environments, in the hectic nature of daily life. By collecting data from hundreds of thousands of patients over time, our real world evidence allows us to measure improvements at both individual and population levels, across all ages, racial and ethnic backgrounds, and socioeconomic conditions. 

Our solutions need to work for everyone, not just for some. This data tells us not only that, but also keeps us grounded and accountable to keep improving and delivering more value. Always.

Health Economics

Better outcomes must also mean sustainable economics. Otherwise, systems cannot scale. 

And our mission is to make healthcare as accessible as running water. We believe AI care can achieve just that.

We study how our solutions modify healthcare utilization and reduce costs—both direct and indirect. Using claims data and third party validated methodology we prove we are delivering value-based care, preventing care escalation and unnecessary spending. More than saving money, we are building a model where better care and lower costs are not in tension.

Building trust, layer by layer

Each of these research areas adds a layer of credibility. Together, they demonstrate that our solutions are grounded in evidence, favorable against best practices, effective at scale for real people, and capable of bending the cost curve.

This is how we usher in a new era in healthcare.

‍