What It Takes to Scale Life Science Software
- Stephanie Roulic
- 9 hours ago
- 4 min read
At Startup Boston Week, Nathan Johnson (CEO, Verne Bio), Piyali Chakraborty (CEO & Cofounder, Ashmi Health), Pradeep Bokinala (CTO, Simbex), Jane Allard (Director of Governance Engineering, GSK) and Ashley Mae Conard (Senior Researcher, Microsoft) tackled one of the most complex challenges in modern innovation: how to build and scale software in life sciences - where messy data, strict regulations, and real human impact collide.
If there was one theme that emerged again and again, it was this: building in this space isn’t just about technology. It’s about trust.
The Reality of “Messy” Biology Data
Life sciences runs on data, but not the clean, structured kind most software teams are used to.
Jane Allard, who leads governance engineering work at GSK, emphasized that designing systems for biology starts with understanding the data itself, its size, structure, and limitations.
“You have to make a plan of what kind of data you’re looking at…and what you’re willing to trade off,” she said, pointing to decisions around cost, latency, and infrastructure.
But even with the right architecture, the reality is unavoidable: biological data is messy.
Ashley (Microsoft Research) underscored that the problem often starts before the software layer.
“This issue of ‘garbage in, garbage out’ is real,” she said, noting that inconsistencies in how data is collected - across labs, experiments, and conditions can introduce bias and noise that systems must account for.
The implication for founders is clear: cleaning and understanding your data isn’t a side task, it’s core to the product itself.
Designing for Humans, Not Just Scientists
While much of the complexity sits in the backend, success often comes down to something much simpler: whether people actually use the product.
For Piyali Chakraborty, founder of Ashmi Health, that means designing around real human behavior, especially in high-stakes, everyday contexts like pregnancy.
“We are really dependent on users finding the time to enter their data,” she explained. “So everything has to be designed around their day-to-day.”
Her team prioritizes user experience in a way that feels more like consumer apps than clinical tools—taking cues from platforms like TikTok and Instagram to drive engagement.
The tradeoff? Less structured data. More cleanup.
But in many cases, that’s the price of building something people will actually use.
Pradeep Bokinala, CTO at Simbex, reinforced the point: even the most advanced system fails if the user doesn’t understand it.
“If the user is not connecting with your application…it’s a loss,” he said.
Validation Isn’t Testing—It’s Trust
In most software, validation means making sure the product works.
In life sciences, it means something much bigger. “When you look at validation… it is building trust,” Bokinala explained.
That trust extends across multiple stakeholders: clinicians, regulators, patients, and partners. And it requires more than just functionality, it requires proof.
Panelists emphasized the importance of documentation, traceability, and reproducibility. Every decision, test, and assumption needs to be recorded not just for internal use, but for audits, regulatory reviews, and long-term credibility.
Allard added that evidence itself becomes part of the product.
“You need to have a way that your evidence is going to be immutable and around for a long time,” she said. For early-stage founders, this creates a unique challenge: building enterprise-level rigor from day one.
AI in Life Sciences: Power With Guardrails
As AI becomes more embedded in life science applications, the stakes only get higher.
Across the panel, there was unanimous agreement: guardrails aren’t optional. “I wouldn’t use any tool that didn’t have guardrails,” one panelist said plainly.
For startups, that often means starting simple: writing down assumptions, defining policies, and implementing human-in-the-loop systems. Chakraborty described how her team incorporates clinicians directly into the loop, ensuring that recommendations are validated before reaching users.
Meanwhile, Allard highlighted how larger organizations are operationalizing governance at scale through tools like policy-as-code, turning human rules into enforceable systems. But regardless of size, the principle is the same: trust must be designed into the system, not added later.
Alignment Is the Hidden Challenge
Beyond the technology, one of the hardest problems in life science software is aligning teams across disciplines. You’re not just building for engineers, you’re working with scientists, clinicians, product teams, and executives, all with different expertise and perspectives.
Chakraborty pointed out that even basic terminology can become a barrier. “Not everybody is at the same literacy level when it comes to AI,” she said. The solution? Simplicity and clarity.
Panelists emphasized the importance of shared language, written documentation, and clear problem definitions, often starting with something as simple as an executive summary. “It’s not real if it’s not written,” one speaker noted.
Building the Right Team
When it comes to hiring, the panel pushed back on the idea that you need massive teams to get started. Instead, focus on the right mix of skills. “You want a subject matter expert in the problem you’re trying to solve,” Allard said, alongside people who can build the underlying data and infrastructure systems.
But perhaps the most important trait isn’t technical…it’s adaptability. Ashley emphasized the importance of a “growth mindset,” especially in a field evolving as quickly as AI and life sciences. “No one can be an expert in everything,” she said. “But if you know how to learn, that’s a huge asset.”
Start Early—or Pay for It Later
One of the most consistent pieces of advice from the panel: do things right from the beginning. Whether it’s data storage, compliance, metadata, or documentation, early decisions compound over time. “All big companies are trying to retrofit this,” Allard said, referring to data standardization and governance. “It’s much easier to start right the first time.”
That includes everything from using secure, compliant cloud infrastructure to properly tagging and organizing data for future use. Because in life sciences, scaling isn’t just about growth. It’s about maintaining integrity, trust, and usability every step of the way.
The Bigger Picture
As AI continues to reshape biology, the opportunity is enormous, but so is the responsibility.
Unlike other sectors, mistakes here don’t just impact users, they impact patients.
And that changes everything.
The founders and operators building in this space aren’t just writing code. They’re building systems that need to be explainable, reliable, and trusted - by everyone who touches them.
Because in life sciences, success isn’t just about innovation, it’s about getting it right.
