AminoChain Natural Search
Role
Product Design
Timeline
3 months | June - August '25
Team
2 designers, 3 engineers, 1 project manager
Problem
Current online biobank marketplaces suffer from outdated search experiences that create friction for users when searching for specimens. How can we streamline this process?
Outcome
I designed a natural search feature and delivered high-fidelity mockups and user flows for the Specimen Center app, using user research to inform my design decisions.
Users struggle to find the samples they need on current online biobank marketplaces.
According to UX audits I conducted on 3 online biobank marketplaces, all suffer from outdated search experiences that create friction for users when searching for specimens. Below is an example of one.
UX audits revealed
slow search experiences
The search processes on these platforms revealed that both a) not surfacing possible diseases and b) having filter conflicts (i.e., anatomic site vs. sample type) slow down the search experience for procurers, who wish to find what they’re looking for within a matter of seconds.
Users don’t use Specimen Center's
search to add samples to their cart
The Specimen Center’s GA Search intended to address the issues present in the search experiences of current online biobank marketplaces; however, users still did not engage according to user research. One hypothesis was that having to click through tabs on the search bar was frustrating to users. Below is the previous search:
User research
results & findings
Analytics tracking data on PostHog revealed the following.
25%
of users who used filters were more likely to add specimens to their inquiry
Despite search being the #3 most commonly used action, no users used search to add specimens to their inquiry.
Users rely on
search to find specimens
Usability tests conducted on 7 participants during the Specimen Center’s Alpha stage revealed the following.
100%
of participants gravitated towards search, despite how these tests excluded the search bar.
During a team discussion on how to improve engagement of the Specimen Center, a major contention was whether to eliminate the search bar or not. However, this data revealed that getting rid of the search bar wasn’t going to fix the problem. Instead, modifying it could potentially do so.
The solution:
an AI natural search feature
Natural Search uses an agentic AI to convert your natural language into a set of filters that are then deployed elastically to retrieve relevant specimens. In addition, it can understand vague queries, including acronyms and lingo. For example, it can synthesize ‘NSCLC tumor+ from adults YTD’ into tumor specimens from non-small cell lung cancer patients in the current year.
“How might a user interact with natural search and filters together or separately?”
3 user flows
guiding the experience
“How might a user interact with natural search and filters together or separately?” is a question my team and I had to consider. I organized this into 3 different flows in Figma to contextualize this:
1
The user enters a query in natural search first, then adjusts filters.
Natural search updates filters. A user can add additional filters or delete ones they don't want to keep.
2
The user uses filters first, then adjusts using natural search.
When the user applies filters and then makes a natural search, the search clears the existing filters.
3
No search results found based on the user’s query.
If a search yields no results, then an empty state must be designed for. See my process for the empty state here.
Designing and prototyping
from scratch
To convey exactly what I wanted natural search to look like to the engineers, I prototyped the search bar in Figma. I designed an elastic (normal) search state, a natural (AI) state, a state with only natural search available, and a loading state. I ensured that my prototype captured how toggling between search results with the arrow keys changed the color of the search button to black (elastic) or green (natural).
