The Situation

Araceli Biosciences had built a high-content imager, Endeavor, which was faster than many of its competitors. However, to bring on more customers, the Endeavor needed an accompanying image analysis application; labs don’t buy an imager that costs hundreds of thousands of dollars without analysis software that works seamlessly with it.

The product team's first attempt to build said analysis application missed the mark in many ways, but the top issue identified by key customers as well as the Sales and Support teams was the lack of ease-of-use, a value proposition that the Endeavor itself stood behind and delivered. Leadership decided to scrap that version and told the team to start from zero.

That decision came with a real cost. As an early-stage company competing against established players, Araceli had only so many attempts to get this right, and killing the first version meant starting over with nothing to show for the time already spent. It also surfaced the actual problem: the team building it had no design leadership at all.

I was hired to rebuild the analysis application from the ground up within six months, and make sure this version finally delivered on what the hardware had promised.

Initial version of the analysis application, scrapped before launch for being too technical for the biologists using it.

The Work

The main challenge when I joined the company wasn't the actual design of the analysis application itself but teaching a team of three software engineers, three data scientists, and a biologist who was also our in-team SME, how to work with a product designer. With no dedicated product manager yet, a few of us, including the data science lead and me, shared the product decisions.

From week one, I ran weekly design reviews where the team could review and ask questions about my proposed workflows and designs, share their own ideas, and have me walk them through why certain design decisions were more appropriate than others given the product requirements. These were technical specialists with strong, well-formed opinions, and design decisions weren't accepted just because I was the designer. Reviews shifted, within a few months, from me defending decisions to the team bringing me their own workflow problems to solve.

The central trade-off was who the application was for. The first version was built like a technical analysis tool that assumed users had engineering and data science knowledge, and I convinced the team that a biologist’s mental model needed to sit at the center. I pushed for software that supports people who go months between sessions, starting with a home page to help them find their footing on return. It’s a familiar pattern in consumer software, and it had to earn its place in ours. I also led the team to a shared language and structure for the whole product, so it could stay flexible and still intuitive for expert users. The team assumed users would read a manual or retain everything from a training session, and I made the case that intuitive was a requirement, not a nice-to-have.

In parallel, I set up monthly design review calls with the Sales and Support teams with the dual intent of keeping them in the loop about product decisions (something that I had discovered they felt removed from) and building confidence in the rebuild. Ahead of launch, I ran the beta program myself, working with the exec team and Sales and Support to shortlist customers, building a testing protocol with specific tasks and surveys, and visiting one beta customer on-site to observe them using the application directly. We shipped the first version of Clairvoyance, analysis software built on machine learning and deep neural networks, at the end of November that year, within six months of beginning from the ground up.

First release of Clairvoyance.

The new year started with me as part of a small "tiger team," led by the VP of Marketing, tasked with building a proof-of-concept for a new product direction set by the CEO and board: letting customers analyze images in real time, as they were being acquired. This product would not only expand the Endeavor platform meaningfully, but would get into a problem left largely untouched by competitors. Starting with storyboarding and moving on to prototyping, I designed the end-to-end concept for the new application. As we refined Clairvoyance with new analysis features, we built a 'shell' MVP for ClaireRT, which we unveiled and demoed as our own internal 'innovation spotlight' at SLAS (Society for Laboratory Automation and Screening), our biggest tradeshow. Customer interest was strong enough that it converted into a purchase order within months and several on-site demos.

Me with ClaireRT MVP at SLAS.

With the successful launch of ClaireRT, I took ownership of product design across Araceli's entire platform. Though each application was powerful and intuitive on its own, together they were inconsistent and disconnected, with no cohesive cross-platform experience. In May 2025, I originated an initiative to unify Araceli's applications into a single, overarching product: one information architecture, one UI language, one taxonomy, and seamless workflows across the platform, while still leaving room for the product portfolio to grow through new applications and add-ons. I led a series of three workshops to bring the core product team, PMs, the dev lead, and our SME, onto the same page and flesh out the details together. The unified app became a recurring topic in board meetings as we discussed the platform's ongoing evolution. In September 2026, we released EndeavorOS, Araceli's single unified application.

EndeavorOS, Araceli Biosciences' unified application.

In 2025, as LLMs became increasingly ubiquitous, I started a personal project surveying people across the company on where they thought LLMs could accelerate drug discovery on our platform. Those conversations surfaced the idea for a Protocol CoAuthor: an agent that would take a scientist’s plain-language description of their samples and generate a starting point for Endeavor’s imaging parameters. With lab-in-the-loop workflows top of mind for the field, I pushed the concept further, closing the loop entirely: the agent would generate the acquisition protocol from the scientist’s description, then use ClaireRT to evaluate the resulting images in real time. With the help of a self-motivated data science intern, I built the Protocol CoAuthor in three months over the summer of 2026. It is now being productized.

I’m also involved in hiring across engineering, QA, and product, including the senior software product manager I now work with.

The Outcome

Customers now consistently cite the software’s usability as a differentiator against other scientific tools, and very few of the feature requests our product and support teams track concern usability; the largest share ask for more analysis capability. Since I joined, the customer base has grown every year, and many customers run more than one Endeavor.

The unified platform, and the products built alongside it, has changed how Araceli's leadership thinks about product strategy, and what role design plays in shaping it. I now co-lead roadmap sessions with the PM overseeing the software portfolio, and present product and design strategy directly to Araceli's board.

I also lead Araceli's Senior Leadership Team, the cross-functional body the CEO convened to drive company-wide initiatives. I set the agenda, hold decision authority within it, and spin up task forces to execute on the priorities that come out of it.

In my three years at the company, design went from something brought in to fix a product that hadn't delivered on its promise to a function that shapes what the company builds next: from a rebuilt application, to a real-time capability, to the unified platform those two now sit inside, and now, to the AI-native tools we're building for what comes after it.