Project 01 / 05
NODO AI
NODO AI
The first AI liquidity engine on Sui needed a product designer. Nobody was going to deposit real money into something they couldn't explain.
Role
Lead Product Designer
Timeline
Q1 to Q2 2025
Scope
0 to 1 Product
Platform
Web App (Sui)
Status
MVP Launched
NODO is an autonomous yield infrastructure protocol. A modular stack of AI agents continuously executes, rebalances, and adapts liquidity positions across Cetus, DeepBook, and Momentum on the Sui blockchain. The design challenge was to make a technically complex financial system feel clear enough that someone holding USDC and 30 seconds of attention could understand what they were getting into.
At a glance
Overview
AI agents managing liquidity across decentralized exchanges on Sui. Vaults, receipt tokens, automated rebalancing. My job was to make all of it feel clear to someone who had never heard of liquidity provisioning.
Challenge
Users distrust vague AI claims. They confuse receipt tokens with rewards. The gap between stated APY and actual returns erodes confidence. Dual-asset deposits block beginners entirely. Every decision had to resolve the tension between transparency and simplicity.
Contribution
I owned the full product design: research program with 24 participants, information architecture, design system, deposit flow redesign, trust framework, and progressive disclosure strategy.
Outcome
Research participants described the redesigned product as significantly clearer than existing DeFi vaults. Single-sided deposit became the default. NDLP comprehension was rebuilt from scratch. Trust signals were placed at every decision point.
Tools & Methods
Figma. Moderated 1:1 interviews. Persona development. Affinity mapping. Concept testing. Prototype walkthroughs. Feature reaction matrix. Opportunity scoring.
Key Themes
Trust in autonomous systems. Progressive disclosure. Yield source transparency. Designing for a spectrum from crypto beginners to institutional allocators. Rebuilding confidence after FTX. Abstracting LP complexity without hiding it.
Context
Why this project existed
In DeFi, earning yield on your crypto means providing liquidity to decentralized exchanges. But managing those positions is technically demanding. You have to choose token pairs, set price ranges, monitor markets, and rebalance constantly. Most users either lose money to impermanent loss or never try in the first place.
NODO set out to automate all of that. AI agents manage liquidity positions across Sui's top DEXs: Cetus, DeepBook, and Momentum. They optimize capital efficiency, mitigate loss, and capture yield from real trading activity. Over $2M in active LP commitments. Over $336M in addressable DEX liquidity on Sui alone.
The market is there. But 63% of users say they would let an autonomous agent manage their funds only if they could understand and trust the system. That trust gap was the design problem.
$536M
DeFAI market cap
$336M
Sui DEX TVL (SAM)
$2M+
Active LP commitments
63%
Users open to AI agents
The Challenge
Defining the design problem
"I just want to put my coins somewhere and earn something. I don't understand why I need to know what liquidity provisioning is."P02, Passive Yield Seeker
Product Challenge
NODO needed to validate its AI vault proposition before a marketing push and convert skeptics in a post-FTX market. The phrase 'AI-powered' had become a red flag. We had to make it mean something again.
UX Challenge
The user spectrum ranged from crypto beginners with $5K portfolios to institutional allocators managing $250K and up. Each segment had different trust thresholds, vocabulary, and mental models. One interface. Six very different people.
Technical Constraints
On-chain vault mechanics required receipt tokens (NDLP), dual-asset deposits, and real-time rebalancing data. Smart contract architecture was non-negotiable. Design had to work within protocol constraints, not the other way around.
Why This Was Hard
No established design patterns exist for AI-managed DeFi vaults. Power users demand full transparency. Beginners need radical simplicity. Competitors had already eroded trust through impermanent loss and hidden fees. We were designing in a trust deficit.
Role
What I owned
Directly Responsible
- End-to-end product design from research to shipped UI
- Information architecture for the vault system
- Design system and component library
- Prototyping, usability testing, and validation
- UX writing across all product surfaces
Collaborated On
- Product strategy and vault roadmap with the founding team
- Technical feasibility scoping with engineering
- Research planning and participant recruitment
- Go-to-market positioning and messaging
Influenced
- Product scope decisions based on research findings
- NDLP token framing and educational strategy
- Feature prioritization through opportunity scoring
- Launch strategy and phased rollout approach
Objectives
What success looked like
User Goals
- Users understand the AI vault value prop within 30 seconds
- First deposit completion exceeds 60% for new users
- Fewer than 20% of users ask what NDLP is after launch
Business Goals
- Grow TVL through better onboarding and visible trust signals
- Position NODO as the first AI liquidity engine on Sui
- Sui-native users make up over 40% of early adopters
Product Goals
- Ship the vault dashboard with TVL, APR, and NDLP tracking
- Default to single-sided deposit for beginner conversion
- Establish scalable patterns for multi-vault deployment in Q3
Trust Goals
- Users understand risk and IL exposure before committing capital
- AI strategy is verifiable through on-chain data, not just words
- Complete fee disclosure including rebalancing costs
Usability Goals
- Over 70% of users identify their net return within 5 seconds
- Beginner task completion exceeds 50% without external help
- Progressive disclosure from summary to full LP decomposition
System Goals
- Composable vault primitives ready for the strategy marketplace
- Full integration with Cetus, DeepBook, and Momentum
- Design system prepared for EVM chain expansion in 2026
Process
How the work unfolded
Understanding the system
Mapped the domain, protocol mechanics, stakeholders, and competitive landscape across Sui and EVM chains.
Identifying friction
Audited existing DeFi vault products, ran heuristic reviews, catalogued every point where users hesitated or abandoned.
Defining principles
Distilled research into six design principles. Every subsequent decision traced back to one of them.
Shaping the structure
Built the information architecture and user flows that organized vault complexity into navigable layers.
Prototyping and iterating
Explored three directions in Figma, tested with users, refined through three rounds of feedback.
Shipping
Aligned with engineering on implementation details and shipped the MVP for Sui mainnet.
Discovery
Understanding before designing
Before touching a screen, I ran a structured research program to understand who we were designing for, what blocked them, and what would make them trust an autonomous system with their capital.
24
Participants interviewed
6
Persona segments identified
45
Minutes per session
14
Insights synthesized
Assumptions We Tested
We went in believing users could articulate the benefit of AI vaults after a 30-second explanation, that most people understood "automated yield" even if LP mechanics confused them, that NDLP would be intuitively understood as a vault share token, and that single-sided deposit would dramatically lower the barrier. Some held up. Others shattered immediately.
Research Methods
Moderated 1:1 remote interviews over Zoom. Figma prototype walkthroughs. Feature reaction matrix scoring. Stimulus testing with live product surfaces. Then a weighted scoring model to prioritize what to build first.
How We Synthesized
All 24 interviews transcribed into a quotes library, tagged by theme and persona. From there I built an insight synthesis matrix — scoring each finding by evidence count, confidence level, and impact on trust, conversion, and retention. That surfaced 14 actionable insights ranked into a prioritized opportunity backlog.
Who We Designed For
Six segments emerged from the research, each with distinct trust thresholds and conversion blockers.
Sui-Native DeFi User
Already active on Cetus and DeepBook
$5K – $100KSkeptical Power User
Evaluates products like an analyst
$25K – $250K+Passive Yield Seeker
Wants yield without managing positions
$5K – $25KStablecoin Safety Optimizer
Preserves capital above all
$10K – $250K+Reward-Driven Explorer
Will try anything with a multiplier
$1K – $10KLP-Confused Beginner
Has a wallet but never provided liquidity
Under $5KTop Pain Points
Mentions across 24 interviews
Top Trust Barriers
What blocks first deposits
"Every DeFi product says 'AI-powered' now. I need to see what the model actually does, not a buzzword on a landing page."
P03 — Skeptical Power User
Insights
What the research revealed
Six findings from 24 interviews that shaped every design decision.
AI is a red flag without proof
“If you can't show me what the model does, 'AI-powered' is just a red flag.”
NDLP is universally misunderstood
“NDLP isn't a reward? Then why does it need to be a token at all?”
Yield opacity triggers Ponzi suspicion
“If you can't tell me where the 8% comes from, I assume it comes from new depositors.”
“It says 12% APY but I only made 7%.
P11, Active DeFi Trader — on APY vs. actual returns
Where did the other 5% go?”
Dual deposit kills beginner conversion
Requiring two tokens to deposit was an immediate abandonment trigger for every user without LP experience. Single-sided deposit became the default — the highest-impact, lowest-effort change in the entire project.
“I have to pick two coins? I only have USDC. This is already too complicated.”
Trust blockers are informational, not technical
Users can deposit. They just won't until they feel they understand the risks. The blockers are about information, not capability.
“I'd put in fifty bucks to test. But I'm not putting savings in unless I see it working for a month.”
Feature Reaction Matrix
Trust levels across persona segments and core product surfaces.
| Feature | Passive | Beginner | Power | Stable | Sui |
|---|---|---|---|---|---|
| AI strategy | |||||
| NDLP token | |||||
| Deposit flow | |||||
| Fee transparency | |||||
| P&L breakdown | |||||
| Strategy docs | |||||
| Audit signals |
Principles
The rules that guided every decision
Simplify the mental model
Reduce the conceptual overhead. Users should understand the system through familiar patterns, not through technical documentation. If someone needs to read a whitepaper to use the product, the product has failed.
Surface trust at key decision points
Proactively answer the questions users have before they ask. Show what is happening, why, and what could go wrong. Trust is built by anticipating doubt, not by ignoring it.
Show complexity only when needed
Layer information progressively. The default experience should be simple. Depth should be available for those who seek it, but never imposed on those who don't.
Prioritize clarity over density
When in doubt, give things more space. Dense interfaces feel powerful, but they erode confidence in users who are already uncertain. Space communicates calm.
Make performance legible
Abstract raw numbers into meaning. Users don't need a spreadsheet of data. They need to know three things: Is this working? Am I safe? What should I do next?
Reduce friction before asking for commitment
Let users understand before they invest. Education precedes action. Confidence precedes commitment. Never ask someone to deposit money into something they can't explain back to you.
Architecture
Structuring the product, not just the screens
The core challenge was taking a system with multiple vault types, receipt tokens, rebalancing logic, fee structures, and risk parameters, and organizing it into something a user could navigate without a tutorial. The architecture had to serve both the beginner who just wants to deposit USDC and the power user who wants to verify every rebalancing decision.
The key architectural decision was to separate the vault system into three information layers. The overview layer shows what you need to decide. The detail layer shows what you need to trust. The technical layer shows what you need to verify. Most users never reach the third layer, and that is by design.
Exploration
Thinking through alternatives
I explored three distinct directions before converging on the final approach. Each had strengths, and the final product borrowed elements from all of them. What mattered was understanding why certain directions were rejected, not just which one won.
Early Concepts
What Was Rejected
Direction A gave power users everything they wanted but overwhelmed beginners on first contact. Direction B held everyone's hand equally, which frustrated experienced users and slowed down the path to deposit. Neither solved the core tension between the six persona segments we had identified.
What Moved Forward
Direction C became the foundation because it respected user agency. Instead of deciding how much information to show, it let users choose their depth. The vault overview is simple. Expanding any section reveals more. The technical layer is always one click away but never in the way.
Solution
The product, explained
Vault Overview
The first thing users see is a clean summary of each vault: the token pair, current APR, total value locked, and a one-line description of what the AI strategy does. No jargon. No walls of metrics. Just enough to decide whether to look closer.
Key decision: We led with net yield after fees, not headline APY. This was a direct response to the research finding that APY creates false expectations.
Single-Sided Deposit
The old flow required users to select two tokens and understand LP pairing. The new flow asks one question: how much USDC do you want to deposit? The protocol handles the rest. Dual-asset deposit still exists for advanced users, but it is no longer the default.
Key decision: This was the highest-impact change in the project. Research showed dual deposit caused immediate abandonment for every beginner segment.
Holdings and P&L
The holdings view leads with one number: your net return. Is your money up or down? That is all most users need. Below that, the P&L waterfall breaks the number apart: gross yield, minus fees, minus impermanent loss, equals your actual return. Power users can expand further into the full LP decomposition.
Key decision: Progressive disclosure. The default is a single number. Each layer adds detail for those who want it.
Yield Attribution
Where does the 8% come from? If we can't answer that clearly, users assume it comes from new depositors. The yield attribution module breaks returns into specific sources: trading fees, LP incentives, and rebalancing gains. Each source links to on-chain data for verification.
Key decision: This module exists because research showed that vague yield attribution actively triggers Ponzi suspicion.
NDLP Education and Strategy Transparency
When a user deposits, they receive NDLP tokens representing their vault position. Because research showed universal confusion about what NDLP is, we added an inline explainer at the exact moment of receipt. For AI strategy transparency, we published both a plain-English summary and a technical brief that power users can verify against on-chain data.
Key decision: Education happens in context, at the moment it is needed. Not in a help center. Not in a tooltip graveyard.
Trust & Clarity
Performance, proof, and action in one screen
The "Your Performance" tab of each vault shows three things at once: how much the vault is earning you, how the AI is allocating across pools, and a deposit panel to add more capital the moment you feel ready.
$412.50 on $5,000
Personal profit and deposited capital are the first two numbers on screen. $412.50 earned, $5,000 committed. Users know exactly where they stand before anything else loads.
Your Rank in the Vault
A distribution chart shows how every depositor in the vault is performing. Each bar is a PnL range. The user’s position is marked clearly. At a glance, you know whether you’re ahead of the crowd or behind it. Social proof through data, not marketing.
Every Depositor, Transparent
The peer comparison table ranks all depositors by PnL. Wallet aliases, deposit amounts, and returns are visible for everyone. Row 23 is highlighted as the user’s position. “Find me” jumps straight to it. Trust comes from showing the numbers that most protocols hide.
Deposit Right Here
The manage liquidity panel sits beside the performance data. After seeing $412.50 in profit, the user can deposit more SUI or USDC without leaving the screen. The context that builds confidence is the same screen that takes the action.
One Token, One Button
The deposit panel defaults to single-token mode. Enter an amount in SUI or USDC, see the estimated NDLP receipt tokens, and hit Deposit. No manual pair splitting, no router selection, no slippage configuration.
26.02% in the Header
APR, total fees earned, total value, and IL ratio are pinned to the vault header. They stay visible across all three tabs: Overview, Your Holdings, and Your Performance. Every tab inherits the same confidence signal.
Outcomes
What changed because of this work
Traction
$2M+
Active LP commitments under management after launch
Confidence
100%
Of shipped screens backed by a research-validated rationale
Coverage
6-in-1
Persona segments served by one progressive disclosure architecture
Delivery
Q2 '25
Shipped MVP on Sui mainnet, on schedule
Qualitative Signals
Research participants consistently described the redesigned product as "significantly clearer" than existing DeFi vault interfaces. The engineering team reported faster implementation due to well-structured design specs. Stakeholders cited the design and research work as a key differentiator in investor conversations.
What Testing Validated
Single-sided deposit removed the most common abandonment trigger for beginners. The NDLP explainer at point of receipt eliminated most comprehension confusion. The P&L waterfall gave users a mental model they could actually hold onto. And the two-phase deposit approach matched exactly how users described their own trust-building process.
Reflection
Looking back, looking forward
What Worked
Leading with research before design saved enormous time. Every major design decision could be traced back to a specific insight with a specific evidence count. The progressive disclosure architecture proved flexible enough to serve all six persona segments without compromising on any of them.
What Remained Unresolved
The reward-driven explorer segment will likely churn when incentive programs end. We designed an education bridge during the reward period, but whether that is enough to convert temporary users into long-term depositors is still unproven. The stablecoin vault, which was the top request from safety-first users, is planned for Sprint 3 but was not yet launched.
What I Would Explore Next
A rebalancing activity log for power users, so they can see every decision the AI made and why. A proper institutional information package with audit reports, legal entity disclosure, and insurance documentation. And deeper work on the strategy marketplace as NODO expands to support third-party vault creators.
What This Changed in My Thinking
This project taught me that trust is not a feature you add at the end. It is the architecture itself. Every information hierarchy decision, every default state, every word choice either builds or erodes the confidence that makes someone willing to commit real money. Designing for financial products is designing for the distance between what people understand and what they need to feel certain about.