layer 01
Hardware input
Sensors, boards, switches, and the physical constraints that make software honest.
FSR pressure board · Basys3 switches
I build hardware-software systems that turn real-world signals into useful feedback.
A systems lab notebook for projects at the edge of code, circuits, robotics, and community service - designed for the messy path between raw data and reliable behavior.
Focused on hardware/software systems, embedded tooling, robotics, and feedback interfaces.
2nd place CPES Hackathon$74 Arduino rehab prototype~20 Hz Web Serial pipeline600 volunteers55% data-fetching complexity reduction200+ Jira tickets
signal-flow/control-system sketchauto · 1/3
what I build
Three layers, one feedback loop.
The portfolio works best when it shows the whole path from physical input to reliable human action.
layer 02
Software interpretation
Calibration, state machines, debouncing, and API boundaries that turn raw events into decisions.
dead zone · release timing · access rules
layer 03
Human-facing feedback
Interfaces that keep the system state visible without forcing people to chase noise.
calmer cue · enriched workflow payloads
Flagship case study · Calibration bench
ReBalance turns unstable sensor input into feedback someone can act on.
- Problem
- Raw pressure signal jitters
- Method
- Calibrate before presenting the cue
- Output
- Shift left / centered / shift right
Jittery raw pressure makes the cue feel nervous, even when the user is close to centered.
Dead-zone and offset logic preserve debug data while presenting a calmer correction.
BLBrian Li · Systems Lab
Working calibration demo
Pressure in. Correction out.
Slider input: calibrated pressure
Move pointer across the board or use the slider.
left footdead zoneright foot
correction thresholdcorrection threshold
Left pressure52%
Right pressure48%
Raw jitter strip
Calibrated score
96CENTEREDOffset +2 · settledMove pointer across the board or use the slider. The raw signal is allowed to be noisy; the cue stays calmer.
How I work when reality gets messy
01Prototype under constraints
02Debug before polishing
03Explain tradeoffs clearly
Field context
Useful feedback for real people.
Hawaii shaped how I think about technology: useful systems should be understandable, local, and human.
I'm a Hawaii-born computer engineering student creating and exploring technology one line, circuit, and prototype at a time.
My projects start with a practical purpose, then move through planning, testing, and repeated refinement until the system communicates clearly.
That shows up in projects where technology has to be useful to people outside the lab: Digital Ready Hawaii seniors learning technology, volunteers coordinating support, patients needing calm feedback, and local communities.
Case files
Artifacts over abstractions.
Preview cards show what was built, what kind of evidence exists, and why the work matters.
Evidence is shown as project facts, roles, artifacts, and process notes - not invented impact numbers.
Hardware-software - Calibration file
Hackathon prototype
sensor -> calibration -> cuecentered
L 52%R 48%
RAW pressure / 20 Hz stream / 4 FSR sensors
ReBalance
A balance feedback system using FSR sensors and Web Serial updates to turn noisy pressure data into a readable correction cue.
Role: Hardware/software developer
FSR sensors, Web Serial stream, calibration, feedback UI
note: learned to separate noisy input from readable feedback
View case studyMarketplace - Trust and verification system
Active MVP foundation
teamsponsorverifylistingreview
Marketplace / verification / Supabase / CI gates
Spontus
A college sports sponsorship marketplace built around verified teams, verified sponsors, structured listings, and application workflows.
Role: Co-founder / product engineer
Domain model, verification workflow, Supabase schema, CI/security gates
note: turned marketplace ambiguity into trust, data, and workflow systems
View case studyFPGA - Digital logic file
Course/lab deployment
clk scan / debounced inputcycle
1wait
2press
3release
4update
Cycles 0-9 like a scanned seven-segment display; hover to scrub WAIT -> PRESS -> RELEASE -> UPDATE.
Flip That Digit
A Basys3 game built around OTTER MCU constraints, seven-segment output, and precise debounce/release handling.
Role: Embedded game logic + hardware deployment
SystemVerilog + Assembly on Basys3 / OTTER MCU
note: learned how release timing shapes reliable state transitions
View case studySystemVerilogAssemblyBasys3
Backend - Operations workflow
Nonprofit production contribution
volunteereventroleapi
CRUD / permissions / MongoDB / coordination
Operation Surf
Backend and data workflows for nonprofit operations supporting volunteer coordination, access-aware CRUD, and API-driven workflows.
Role: Backend/API workflow contributor
MongoDB/Mongoose, Next.js, API workflows, permissions
note: learned how backend structure affects real-world coordination
View case studyvolunteer opsMongoDBaccess-aware CRUD
IDE extension - Learning loop
Hackathon prototype
explainquizgradeadapt
local IDE loop / validated output / adaptive engine
Vybe Tutor
A local-first VS Code extension that turns AI-generated code into learning opportunities: explanations, quizzes, and an adaptive engine.
Role: Contributor - adaptive engine & MCP integrations
Gemini-driven explanations, Zod-validated outputs, deterministic adaptive engine, MCP doc-enricher
note: privacy-first grading and deterministic difficulty transitions
View case studyTypeScriptVS Code Extension
Debug log / Failure notes
Raw engineering process stays visible.
A teaser for writing that reads like a lab notebook, not a polished victory lap.
Entry 001 · ReBalance calibration failure
failurecalibrationcalmer cue
When the signal got noisy, I stopped tuning the UI and went back to the measurement model.
01Context
Raw pressure values drifted enough to make the correction cue feel nervous.
02Decision
The fix was not more visual polish — it was calibration: separate noisy input from readable output.
03Next test
Next pass: tune dead zone, expose raw strip, keep the user cue calm.