PRODUCT · Aug 2022 – May 2023
Sole designer of mrgnlend - the core lending experience that scaled to hundreds of millions in TVL
Context
Borrowing against volatile assets made every action a risk decision, yet DeFi interfaces often exposed protocol mechanics without helping users understand what would happen to their position.
Product decision
Make risk a first-class interaction. I organized lending, borrowing, portfolio state, and liquidation around a continuously legible health factor, then carried that model from research and wireframes into the shipped product.
Outcome & attribution
mrgnlend launched on that foundation. I left before the points program and the market expansion that followed; the product later scaled to hundreds of millions in TVL and peaked above $800M, so that growth is context - not a claim of sole causation.
Research through shipped core experience.
Reached after my departure; included with attribution.
Prototype, UI system, and production states are linked.
Key decisions
Borrowing against volatile assets meant every action changed a user's liquidation exposure. I made Health Factor and live LTV persistent parts of the interaction rather than information users had to hunt for after acting.
The product needed to support first deposits, active positions, market comparison, and position management without splitting one account across competing mental models. A shared asset-row and account system kept those states connected.
I worked from research and flows into empty, connected, funded, borrowing, warning, and approval states, then iterated the core rows and controls with engineering until the interaction held together in production.
Black Box, Superstake, and Omni explored different products and interaction models, but each reused the same emphasis on legible state, explicit risk, and a coherent mrgn product world rather than becoming disconnected experiments.
Project chronology
The switcher below separates distinct workstreams; this chronology shows how the risk model and interaction language moved between them during the role.
01 / 05
As the sole product designer, I took mrgnlend from research and risk-model definition through UX architecture, interaction design, component iteration, and the shipped borrowing, lending, portfolio, and position-management experience. I left before the points program and major growth phase; the product foundation remained in place as the protocol later scaled to hundreds of millions in TVL, peaking above $800M after my departure.
01
mrgnlend brought supplying, borrowing, account health, and position state into one responsive surface. The product film establishes the complete interaction before the case study moves backward through the decisions that shaped it.
02
The interface made account value, supplied assets, borrowed assets, available liquidity, and Health Factor readable together. Users could understand their position before choosing the next action.
03
A shared row pattern let users compare markets, switch intent, and act without relearning the page. The surrounding account state remained stable while the action changed.
04
The LTV control updated the consequence of an action while the user was still deciding. Risk was not a warning added at confirmation; it was part of the primary interaction.
05
I mapped how experienced and newer DeFi users monitored positions, compared yield, interpreted account health, and recovered from risk. That work defined the product hierarchy before visual design began.
06
This presentation documents the thinking behind mrgnlend: the product context, risk-first model, user flows, interface iterations, and the path from early research to a shipped borrowing and lending system.
07
The system moved through structural wireframes, product-model debates, and increasingly resolved interface states. Each round clarified what belonged in the account summary, market rows, and action layer.
08
The empty state taught the model and made the first action obvious. Once funded, the same structure expanded into a dense account view without changing the underlying hierarchy.
09
Lend and Borrow rows carried market information, wallet state, position state, and the primary action. Five focused rounds resolved density, hierarchy, and behavior before the pattern scaled across assets.
10
The launch pages reduced a complex protocol to an understandable promise while preserving the same visual world as the application.
11
The working prototype behind mrgnlend - research, flows, interface states, and risk interactions from the core borrow/lend product.