Mechanical and aerospace engineering. Everything on the mat, as a list.
FTC DECODE season
Turret and Vision Detection System
A complete study into the mechanics of a turret that can independently control yaw, tilt, and launch distance for a spherical projectile.
Role
Mechanical design and programming
Tools
Fusion 360, Limelight, Java
Range
Projectiles up to 25 ft
Pitch change
0.25 s or less
0°Drag to rotate
The final design. Drag across it, or use the arrow keys, to turn it all the way around.
Revisions
The turret system was built over the course of several months, and dozens of versions. Select a revision to see it.
Rev / description
A functional, ready-to-make turret that can aim and fire projectiles up to 25ft.
How it works
Four details from the final design. Move across any of the small views to look around it up close.
Y
Detail YScale 5:2
Yaw
The turret rotates on a large ring gear at its base, with a roughly 3:1 ratio. A lower adapter plate allows either a servo or a motor to drive the system.
P
Detail PScale 2:1
Pitch
Curved gear racks on each side tilt the shooter to change the launch angle, fixed to a high-speed servo. The turret can fully adjust its angle in .25s or less.
F
Detail FScale 2:1
Flywheel
A PID loop sets the flywheel RPM based on a data feed from the vision system providing target distance into a regression model made with real-life training data. A belted system unites two high-rpm motors to deliver optimal torque while maintaining the ability to quickly change speeds.
V
Detail VScale 5:2
Vision
A Limelight machine learning camera on the front finds the target and measures its distance using trigonometry. The resulting data feeds the flywheel's PID loop and the turret yaw control to perfectly track a target in real time.
The math behind the shot
For my IB Math internal assessment, I worked out the launch angle a flywheel shooter needs to hit the goal from anywhere on the field, so the robot can aim from live position data instead of angles guessed from test shots.
The paper models a hooded shooter for the 2026 FRC game: a 6 in foam ball, a goal 6 ft off the ground, and a 3 in flywheel spinning at 2,500 rpm, which works out to about 10 m/s at launch. Eliminating time from the projectile equations leaves a quadratic in tan θ, so every distance has two answers: a low, fast arc and a high, lobbed one.
θ = tan⁻¹ [ ( v² ± √( v⁴ − g( g x² + 2 v² ( y − h₀ ) ) ) ) ÷ g x ]x is the distance to the goal, y the goal height, h₀ the launch height, v the launch speed.
Low arc –High arc –
Drag the distance. The arc the robot would use is drawn solid.No air resistance or spin, same as the paper
Below 5 meters, the low arc clips the front of the goal, so my proposed solution switches to the high arc inside that range and uses the low arc beyond it, where the shorter flight time makes shots more consistent. The next step is testing it on a real shooter and tuning the model with the data.
ProjectTurret and vision system
Drawn byJ. Noah
RevD
Sheet1 of 8
2024 Congressional App Challenge winner, WA-04
Canari
Agriculture insights for everyone. A free iOS app for reporting crop pests, disease, and climate damage, built for everyone from hobbyist gardeners to industrial farmers.
Role
Design and development
Built with
Swift, SwiftUI, Firebase
Started
Spring 2024
Today
In use at a local farm
Record a report
See reports nearby
Read the details
“Every crop saved is a step towards reducing food poverty.”
From my interview with the Tri-City Herald
Why I built it
My mom works in agriculture, and one night over dinner in spring 2024 she told me she wished there was an efficient, customizable way to track and solve issues across a farm. Her farm still relied on clipboards to manage information.
Canari started as farm management software for companies. After pausing development to focus on school, I realized it could be much bigger: a free tool that helps reduce food waste and works toward a future without food poverty. Canari exists today as private software developed for a local farm in my region, mostly utilized for tracking empirical data in fruit.
One database, two kinds of users
Hobbyist gardeners
See general trends in pests, disease, and climate damage around your location, based on anonymized reports from farms.
Farms
Report issues internally, create presets specific to your organization, and use crowdsourced reports from hobbyists to help optimize yield.
How it's built
Swift and SwiftUI
Canari is built entirely in Swift, Apple's programming language, for ease of use and compatibility with the Apple ecosystem.
Firebase
Google's Firebase handles all of the databases, along with user accounts.
Live presets
Presets and reports load dynamically on launch, so I can update the database without users having to update the app.
Privacy
Reports are geolocated to within about 8 feet, then the sensitive data is scrambled before it reaches the shared database, so farms can see regional trends without exposing their own operations.
Testing it in the field
Recognition alone doesn't create impact, so I took Canari onto a farm. During my software internship at a local fruit company, I gave the farm access to the app in exchange for live testing. Many of the workers I tested with don't speak English, and their feedback led me to change the app's language, visuals, and logic.
I also learned that one app is too static for agriculture. Every farm tracks different things, so Canari is becoming a customizable platform: each farm builds out the app for its own needs, while anonymized data still flows into one shared database.
Timeline
Spring 2024A conversation with my mom about her farm's clipboards.
May 2024Started building Canari while learning Swift.
Fall 2024My programming teacher encouraged me to finish it, so I rebuilt the app from the ground up in about a week for the Congressional App Challenge.
December 2024Chosen as the winner for Washington's 4th District. That year's challenge drew 12,682 students across 382 congressional districts, and Canari was selected for display in the U.S. Capitol and on House.gov.
AfterwardMet with Rep. Dan Newhouse in Washington, D.C. to talk about how the project could be advanced.
Summer 2025Live testing on a local farm, and a redesign based on what the workers told me.
Coverage
Rep. Dan NewhouseNewhouse Announces Winner of the 2024 Congressional App Challenge
Tri-City HeraldTri-City student honored by Congress for a tracking app for farmers
Government TechnologyWashington Student Builds App to Track Food Supply Issues
ProjectCanari
Drawn byJ. Noah
Version1.0
Sheet2 of 8
Personal project, in progress
Electric Honda CT90
Converting a 1968 Honda CT90 into an electric motorcycle. I restored the frame from scratch, designed the motor mounts in Fusion 360, and I'm now fitting a BBS02 motor system.
Role
Restoration, CAD, and fabrication
Tools
Fusion 360, photogrammetry, 3D printing
Drive
BBS02 motor system
Status
Mounting the motor
The build so far
FoundA 1968 Honda Trail that had been sitting for years with a blown engine. Yellow, and in bad condition.
RestoredI took the bike apart, restored the frame from scratch, and had it powder coated blue.
DesignedFitting a modern motor to a 1968 frame meant designing parts around mounting points that were never meant for it. I used photogrammetry and reference photos in Fusion 360 to work out the frame's geometry, then created a dozen mount designs.
PrintedPrinted the adapter plates and test-fit them on the real frame.
NowI chose a BBS02 motor system and I'm now mounting it to the restored frame, checking alignment as each part goes on.
The motor mount
The mount carries the BBS02 and its chainring, and bolts to the frame's original mounting points. Rendered in Fusion 360.
AIsometricBDrive sideCBack side
What's next
Once the motor is mounted, I'm working out the finer details: wiring up the original gauges, and adding solar so the bike can recharge off-grid. The goal is to ride it around campus this spring.
ProjectElectric Honda CT90
Drawn byJ. Noah
StatusIn progress
Sheet3 of 8
Arduino microcontroller certified
Electronics
I've been doing electronics for five years, since my freshman year of high school. Most of that work comes from wiring competition robots. I also converted an old gate to passcode access, and I completed a year-long course in digital electronics.
Experience
Five years, since freshman year
Certification
Arduino microcontrollers
Coursework
Digital electronics, one year
Tools
Arduino, Multisim
A year of digital electronics
Circuits
Building and analyzing circuits.
Logic gates
Combining gates into working logic.
Truth tables
Mapping every input to its output.
Multisim
Designing and simulating circuits before building them.
Robot wiring
I've done a lot of the wiring on our competition robots, connecting the control electronics, motors, and dozens of sensor types. I also taught younger students on the team how to wire.
Control electronics laid out on a robot.A competition robot mid-build.
Converting a gate to passcode access
I converted an old gate so it could only be opened with a passcode. That meant wiring into the gate's existing control panel and installing a receiver to connect the two.
The receiver, wired into the gate's original control board.
ProjectElectronics
Drawn byJ. Noah
CertArduino
Sheet4 of 8
Lion's Byte Robotics, Kennewick High School
Leadership
I cofounded my high school's robotics team as a sophomore and became its captain and head of design and engineering. We went straight into the FIRST Robotics Competition, and made it to the world championship in our rookie year.
Role
Cofounder, captain, head of design and engineering
Program
FIRST Robotics Competition
Seasons
Three, sophomore through senior year
The banner wall
2024Rookie All-StarClackamas
2024Rookie Engineering InspirationYakima Sundome
2024Rookie All-StarPNW District Championship
2024World ChampionshipHouston
CaptainLeague ChampionsEastern Washington
CaptainChampionsHanford Engineering Competition
Season by season
2024Sophomore year, Crescendo
Rookie year, and a trip to Worlds
Instead of starting in an easier league, we went straight into FRC. After weeks of work and plenty of setbacks, we won Rookie All-Star at Clackamas and at the PNW District Championship, and Rookie Engineering Inspiration at the Yakima Sundome. We fundraised to bring our small team to the world championship in Houston. Seeing the design work there set the bar I've been chasing since.
2025Junior year, Reefscape
A season that didn't go to plan
At our first competition, we found out the narrow base we'd designed for agility made our tall robot wobble and tip at speed. New wheel material, weights on the base, and driving slower all made it sluggish, and we didn't have the time or money to rebuild. So we cut away a large section of the structure and played defense for the rest of the season, and we had more fun than we'd had all year. I learned there's enormous value in small victories.
2026Senior year
Captain, lead engineer, league champions
As captain and lead engineer, I designed mechanisms and wrote control code, including the aiming math for our shooter. We won the Eastern Washington league championship, and I captained our team to the win at the Hanford Engineering Competition.
Building the team
Leading a new team meant teaching as much as building. I mentored younger students in CAD, wiring, and programming so they could contribute to the robot on their own, and I worked on the robot almost every day after school.
One of our robots in the shop.
Outside robotics
Student government
Senior class president; organized large events, including the senior class party
Service
National Honor Society alum
ProjectLeadership
Drawn byJ. Noah
TeamLion's Byte
Sheet5 of 8
Software development internship, Taggares Fruit Company
Claude agents for agriculture
I build AI agents on Claude that connect straight to a company's own data, so anyone on the team can ask a question in plain language instead of digging through files and spreadsheets.
Where
Taggares Fruit Company, Burbank, WA
When
June 2025 to August 2026
Built with
Claude, Markdown skill files, internal APIs
Status
Deployed internally
How the agent is put together
InAn employee's question
AgentClaudeInstructions and skills written as Markdown files
ToolsCompany filesInternal APIsLive business data
OutAnswers, charts, and graphics
One place to ask, instead of a dozen places to look
What I built
An internal Claude agent. It has access to company files, internal APIs, live business data, and custom skills for generating charts and graphics, so employees can query company data through one interface.
Agents defined in Markdown. Each agent's instructions and skills live in Markdown files, with direct API connections to the farm's data.
Data apps for the orchard. Prototypes to track apple chemical levels, pruning data, and other measurements of fruit production and maturity.
A rebuilt time off system. Automated PTO and PSL requests, approvals, email notifications, and syncing with a shared company calendar.
ProjectClaude agents
Drawn byJ. Noah
StatusDeployed
Sheet6 of 8
Ongoing experiment
Kalshi paper-trading agent
An agent that reads celebrity news every minute, argues with itself about the odds, and trades Kalshi's prediction markets on paper. No real money, just a test of whether the strategy holds up.
Markets
Celebrity news
Cycle
Every minute
Brain
Claude, plus a council of personalities
Money
Paper only
One minute, start to finish
1
Scan
Pulls the latest from celebrity news sites.
2
Remember
Updates a Markdown summary for each topic.
3
Estimate
Forms its own idea of the odds.
4
Debate
A council of personalities pushes back.
5
Buy
Sizes orders to the contracts actually available.
6
Manage
Rebalances on fresh prices and fresh news.
Repeats every minute
The council
One opinion is easy to fool, so the agent doesn't trade on its first guess. It hands its estimate to a council of AI personalities, each built to read the same news differently, and the odds that come out of the debate are what it trades on.
InThe agent's estimate
PersonaThe skepticLooks for reasons the story is noise and the market already has it right.
PersonaThe optimistLooks for reasons the news will actually play out.
PersonaAnd othersMore voices with their own angles on the same facts.
OutThe council's odds
Memory in Markdown
Everything the agent learns is written down. Each topic gets its own Markdown file with a running summary of recent events, so every minute the agent starts from what it already knows about that topic instead of from scratch.
The same idea runs through my Claude agents for agriculture: plain Markdown files are easy for the model to read, easy for me to check, and easy to keep up to date.
topics/topic-c.md
topics/topic-b.md
topics/topic-a.mdSummary of recent eventsCurrent read
Buying and holding
The buy algorithm
Kelly sizing says how much the edge is worth. The buy algorithm then adjusts the order to the contracts actually available at the market buy price, so it never plans around liquidity that isn't there.
Keeping the portfolio
Every minute, market data refreshes. The agent recomputes Kelly for each position with the latest prices and whatever news just broke, then trims, adds, or holds.
Time arbitrage
The edge it's hunting: the gap between when news breaks and when a market's price catches up to it.
Paper only
Every trade is simulated. The point is to find out whether the strategy holds up before any real money is involved.
Sizing a position
A Kalshi contract pays $1 if the event happens. If a contract costs c and the agent thinks the real probability is p, the Kelly fraction of the bankroll to put in is (p − c) ÷ (1 − c).
Edge 15¢Kelly stake 25%Half Kelly 12.5%
No edge, no bet. The stake drops to zero when the price is at or above the agent's probability.
ProjectKalshi agent
Drawn byJ. Noah
MoneyPaper
Sheet7 of 8
Mechanical and aerospace engineering
Jeffrey Noah
I'm studying Mechanical and Aerospace Engineering at Case Western Reserve University, class of 2030. I've designed and programmed robots, built an agriculture app that won the Congressional App Challenge, and I'm converting a classic Honda motorcycle into an electric vehicle.
Case Western Reserve University, Mechanical and Aerospace Engineering, 2026 to present. Kennewick High School, full IB Diploma with academic honors.
Experience
Software Development Intern, Taggares Fruit Company Burbank, WA. June 2025 to August 2026
Rebuilt the company's PTO and PSL management system, automating time off requests, approvals, email notifications, and syncing with a shared company calendar.
Developed and deployed an internal Claude agent with access to company files, internal APIs, live business data, and custom skills for charts and graphics, so employees can query company data in one place.
Built data analysis apps and prototypes to track apple chemical levels, pruning data, and other measurements of fruit production and maturity.
