Case study
Airframe AI
I joined Airframe as one of its earliest engineers and helped build the core product from scratch. As my role grew from frontend engineering into product ownership, I led the design and development of experiences that made AI software easier to discover, evaluate and adopt, from universal search and expert matching to a new approach to retention.
Client
Airframe AI (founding team)
Year
2025 — 2026
Role
Product, Design, and Engineering
Platforms
Web app, newsletter
Outcome
Built the core product behind a $4M seed round, then led search, expert matching and retention bets end to end.

Choosing AI software is noisy, and it only happens a few times a year.
There are thousands of AI tools, and the leaders choosing between them are buried in hype and vendor noise. Airframe set out to be the place they come to cut through it: compare products, read real use cases and talk to experts who have used them.
The catch: companies only buy software a few times a year. The product had to be useful enough to win the buying moment, and interesting enough to come back to in between.
Second hire, many hats.
I joined Airframe in February 2025 as one of its earliest engineers and owned the frontend architecture, working directly with the co-founders to build the core product from scratch, the product Airframe raised its $4M seed round on.
As the company grew, my role expanded beyond engineering. I led projects end to end, from product strategy to UX design, technical architecture and implementation. I wrote PRDs, prioritized features, designed workflows and shipped them myself. Being both the designer and the engineer meant I could move quickly from an idea to a working product, while understanding the tradeoffs behind every decision.
Study, sketch, test, then build.
Every project started with how people actually make decisions. I studied products that solve similar problems, from LinkedIn and Spotify to GLG and Morning Brew, and sketched widely before committing, including 40+ low-fidelity layouts for the news page alone.
Then I tested the cheapest version first, with internal users, a prototype or an email, before building the real thing.
Easier to search, easier to trust, worth coming back to.
Airframe had three parts, products, experts and RFPs, that each lived in their own silo. I led three bets to pull them together: one search for everything, a trustworthy way to reach the right expert without leaving the product, and a reason to come back between purchases.
Search once, find everything.
3 → 1
separate searches for products, experts and RFPs, merged into one
The problem
Products, experts and RFPs each had their own search, so finding what you needed meant knowing where to look and searching three times. People also wanted to search by category and industry, not just by exact name.
The call
Product: unify discovery so people don’t have to know where to search. One search bar returns products, experts and RFPs grouped by type, with a scoped search kept for power users. Technical: I built it on Typesense and structured the search data on the frontend to work around backend schema limits, so search could keep up as the data grew.
The tradeoff
I started with LinkedIn-style filters, but as the list of filters grew, they crowded the results. A simpler collapsible sidebar scaled better, even if it was less flashy.
What happened
Universal search shipped with analytics tracking what people searched for and which categories they opened.
The best design isn’t the most sophisticated one. It’s the one that still works when the product doubles.

How do you make expert introductions feel trustworthy before you can automate them?
14
internal users tested two request flows, and the prioritized list won
The problem
Experts lived in a separate directory, disconnected from the products and RFPs people were researching, so introductions happened outside Airframe at the moment people most needed guidance.
The call
I chose a human-assisted experience over automated matching. Expert availability was uncertain and trust mattered, so people pick experts from inside product pages and the RFP flow, and Airframe’s concierge team makes the introduction. If someone isn’t available, the product suggests similar experts so no one has to start over.
The tradeoff
A concierge model doesn’t scale like self-serve booking. I bet that getting the right expert mattered more than getting the fastest one, and that automation could come once the matches were proven.
What happened
I tested two request flows with 14 internal users: one expert at a time, or a prioritized list. The list won clearly. People hesitated before requesting a single expert, worried about using up their one shot. Letting them build a list removed that pressure and gave the concierge team backup options for every request.
Don’t make people use up their one shot.

How do you build a habit around something people only do twice a year?
20
buyers and recommenders in the external newsletter pilot
The problem
Buying happens a few times a year, so Airframe needed a reason to visit in between. Leaders were drowning in AI content but short on credible, quantified examples they could bring to a meeting.
The call
I designed News Desk, a daily feed of real AI use cases with measurable results. Instead of building it in the app first, I launched it as a newsletter: five hand-picked use cases a day, first to the team, then to 20 buyers and recommenders from Airframe’s expert network.
The tradeoff
Testing through email let me validate the content and format before investing in a new product surface. The tradeoff was that email engagement wouldn’t necessarily translate into people returning to Airframe itself.
What happened
Short summaries with a hard number, like a 40% drop in support tickets, got the most engagement. Two small A/B tests pointed the same way: titles written for the reader’s role did better than generic ones, and bullet summaries did better than paragraphs. Those findings shaped the design of the in-app News page.
Retention isn’t a feature. It’s a habit you have to earn.

Start lean, then earn the habit.
Each bet taught me to test the smallest version first: a sketch before a build, 14 internal users before a launch, an email before an app.
And the features that worked best didn’t ask people to go somewhere new. They showed up inside work people were already doing, whether that was researching a product, writing an RFP or checking their inbox.
$4M
seed round Airframe raised on the core product I helped build
0 → 1
frontend architecture, built from scratch
3
products led end to end: universal search, Expert Connect and News Desk
Onboarding, the homepage and RFP workflows.
Beyond the three bets, I designed and built Airframe’s multilayered onboarding, which helped new users understand a complex, unfamiliar platform, along with the homepage and the RFP workflows.