2023Agri-Tech / SaaS
Aotomatic
Aquaculture data entry — 16 steps down to 4. Same backend rules. Completely different experience.
- UI/UX Design
Designed for cleanroom operators, strict accuracy, and zero room for ambiguity.
I joined this project three years into development, with no room for visual guesswork. My role is to translate dense hardware specs into an operator-facing interface that stays clear, safe, and aligned with machine behavior.
[01]IMPACT & SUMMARY_
HMI for a wafer grinding machine, designed for gloved operators, strict accuracy, and a zero-error environment.
Project Status
Ongoing
Active Product Designer
Year-3 Entry
Onboarded into a mature, multi-year R&D cycle
Strict NDA
Classified visual assets & core hardware IP
Project Status
Ongoing
Active Product Designer
Year-3 Entry
Onboarded into a mature, multi-year R&D cycle
Strict NDA
Classified visual assets & core hardware IP
I joined this project three years after it started — with no brief, no UX docs, and a requirement that nothing could be wrong.
I’m working inside a three-vendor software setup for a wafer grinding machine project that was already deep into development when I joined. My role is to turn years of hardware logic and engineering constraints into an interface the frontend teams can build and operators can trust.
Executive Summary
- I’m responsible for translating complex hardware specifications into a clean, operator-grade interface. My work focuses on making machine states legible, actions unambiguous, and interaction safe for operators in cleanroom conditions.
- The HMI passed hardware validation and keeps machine logic in sync with what operators see and do on screen.
[02]THE PROBLEM_
THE PROBLEM
INHERITING 3 YEARS OF
HARDWARE LOGIC.
Joining a deep-tech hardware project in its third year meant coming into a machine that already worked, with years of engineering decisions baked in. Instead of a blank canvas, I inherited a dense pile of specs and Confluence pages. A visual mistake here isn’t just an annoyance—it can waste wafers and time on expensive equipment.
The hardware core is controlled by a Japanese semiconductor manufacturer, while software is split across three Vietnamese vendors under strict NDA. I had to design precise machine state visualizations without ever having direct, open access to the hardware itself.
[03]DISCOVERY & APPROACH_
DISCOVERY & APPROACH
DOMAIN KNOWLEDGE OVER
DESIGN TRENDS.
Under heavy NDA, standard UX research methods are impossible. Discovery meant spending my initial months decrypting years of engineering documents before opening a design tool.
Working from specs, not interviews
- Sensor & State Mapping: Analyzed dense Confluence manuals, sensor layouts, and backend status codes to map out every possible machine state that needed to be represented on the interface.
- The Validation Loop: Set up a simple but strict loop: read spec → propose UI state → review with hardware engineers → only then hand off to frontend. Nothing went to dev without hardware sign-off.
- Cross-Vendor Alignment: Kept the UI logic aligned across multiple outsourced software teams so the final build would match the manufacturer’s expectations of how the machine behaved.
[04]Key Design Decisions_
Key Design Decisions
DESIGNED FOR CLEANROOM USE.
Glove-Optimized UI
Scaled tap targets beyond 50px and simplified layouts so operators in cleanroom gloves could still hit the right control without hesitation.
Hierarchical Alarms
Turned raw hardware codes and sensor signals into human-readable, prioritized alarms. Operators see what’s wrong, how severe it is, and what to do next, instead of just a flashing error code.
React-Ready Specs
Documented clear component states and data expectations so the frontend teams could wire machine logic to UI without guessing how a state should look.
Glove-Optimized UI
Scaled tap targets beyond 50px and simplified layouts so operators in cleanroom gloves could still hit the right control without hesitation.
Hierarchical Alarms
Turned raw hardware codes and sensor signals into human-readable, prioritized alarms. Operators see what’s wrong, how severe it is, and what to do next, instead of just a flashing error code.
React-Ready Specs
Documented clear component states and data expectations so the frontend teams could wire machine logic to UI without guessing how a state should look.
[05]Reflection & Lessons Learned_
Reflection & Lessons Learned
01
Precision over creativity: In industrial HMI, "innovative" UI is dangerous. Familiarity and accuracy are the only metrics that matter. The screen must safely reflect the machine's state.
02
NDA is a design constraint: when you can’t show visuals or talk about the product openly, you’re forced to explain your value through reasoning, structure, and process instead of pixels.
03
Design as Translation: The true value of a product designer in deep tech isn't drawing interfaces; it's translating years of Confluence specs and strict hardware constraints into operational safety.
