CHELBIAN
// Kelton Chelbian  ·  Auburn, IL

I Build
Things

Automotive diagnostics. Embedded systems. Robotics. Software. If it has wires, an engine, or a compiler, I've probably taken it apart.

SACA C-101
SACA C-201
SACA C-202
OSHA Certified
FRC Team 4156
Scroll to explore
01

About Me

I'm a high school senior with a hands-on technical mind and a creative drive that pushes me past surface-level answers. I don't approach problems academically — I approach them the way a mechanic or engineer does:

What is it actually doing? What's really causing that? How do I fix it?

My interests span automotive diagnostics and performance tuning, electronics and power systems, computer vision and robotics, and software development — and I run projects across all of them at the same time.

I hold SACA certifications in manufacturing and automation, OSHA certification, and competed on FRC Team 4156 as a developer and builder. I learn fastest by building something real, breaking it, and debugging my way back.

3
SACA Certs — C-101, C-201, C-202
OSHA
Safety Certified
4+
Technical Domains Active Simultaneously
FRC
Team 4156 — Developer & Builder
02

Skills

Automotive
ECU diagnostics SCT X4 tuning Fuel trims / AFR Exhaust fabrication O2 sensor analysis Bodywork / rust repair
Electronics & Power
Li-ion battery systems PWM motor control BMS / CC-CV charging Arduino DC power systems 18650 cell harvesting
Software
Python Java (WPILib) OpenSCAD / 3D design NetworkTables Computer vision (YOLO) Linux CLI
Robotics
FRC Team 4156 Swerve drive systems Limelight / AprilTag Jetson / NanoOWL Phoenix 6
PC / Networking
Windows & Linux Throttling diagnostics Pi-hole / DHCP VPN passthrough Audio routing
Certified
SACA C-101 SACA C-201 SACA C-202 OSHA Manufacturing fundamentals Industrial automation
03

Projects

FRC Team 4156
Ball Detection & Auto-Aim
Robotics · 2025–2026
WPILib Java · Python · Jetson Orin

Built a dual-vision system for Team 4156's competition robot. The primary pipeline ran NanoOWL on a Jetson Orin coprocessor — an open-vocabulary object detector that identified yellow FUEL balls in real time and streamed bounding box data over NetworkTables to the roboRIO.

When the Jetson had CSI camera issues during competition, I integrated a Limelight 3 as a drop-in replacement, writing a color-threshold pipeline for ball detection and a separate AprilTag pipeline for Hub targeting and autonomous distance calculation using the ty angle offset.

Debugged GPIO pin conflicts when the tilt stepper motor failed to initialize — traced to GPIO.setmode() not being called before output, and resolved with a proper cleanup/init sequence. Built the full Java robot architecture on WPILib + Phoenix 6 including a swerve drive base.

JETSON NanoOWL YOLO detect LIMELIGHT 3 AprilTag + Color NETWORK TABLES 10.41.56.2 robo RIO WPILib bbox data aim cmd CSI camera
NetworkTables · Java vision command
double ty = LimelightHelpers
  .getTY("limelight");
double dist = (HUB_HEIGHT - CAM_HEIGHT)
  / Math.tan(Math.toRadians(
      CAM_ANGLE + ty));
// feed to shooter RPM lookup table
double rpm = distToRPM(dist);
18650 Battery Pack
& PWM Water Gun
Electronics · 2026
Li-ion · PWM · BMS

Built a 7.4V series Li-ion pack by harvesting Samsung 18650 cells from WORX 20V PowerShare drill packs — each pack yields 5 cells, 2 used in series for the gun, 3 spare. Identified that WORX "Nitro" and "ProPlus" labeled packs use non-standard cell formats and avoided those.

Wired a PWM motor speed controller (5–10A rated) directly from the pack to the gun motor, bypassing the boost converter that was browning out at the motor's 3–5A draw. The boost converter was rerouted to charging duty only, running CC/CV into the pack with manual monitoring at 7.7V.

Determined empirically that 8V is the sweet spot — above that, diminishing returns on speed with increasing heat. Below 7.4V, pressure drops noticeably. The nozzle diameter, not voltage, turned out to be the real range limiter.

18650 3.7V Samsung + 18650 3.7V Samsung 2S series = 7.4V 14 AWG 15A fuse PWM CTRL 5-10A rated PUMP MOTOR @8V opt
Specs
CellsSamsung 18650 (from WORX 20V pack)
Config2S series — 7.4V nominal, 8.4V max
ControllerPWM, 5–10A, variable speed
Fuse15–20A inline blade on positive
Sweet spot8V operating — empirically tested
Wire14 AWG minimum
WORX PowerShare
Battery Adapter
3D Design · 2026
OpenSCAD · PETG/ABS

Designed a fully parametric 3D-printable adapter that clicks onto any WORX 20V PowerShare battery the same way a tool would, exposing clean positive and negative terminals for external wiring. The WORX connector has 5 pins — only the 2 power pins matter; temp, voltage sense, and ID pins float.

Went through three full design iterations in OpenSCAD. V1 established the dovetail rail channel and contact geometry. V2 added a cantilever latch arm and spade connector housings. V3 matched a compact rectangular block form factor with top-face contact holes, a side-mounted cantilever latch with downward hook, bottom-accessible spade pockets, and internal wire routing channels.

Key design parameter: a tunable FIT variable (default 0.2mm) adjusts rail channel clearance for different printers. Material must be PETG or ABS — PLA will deform from battery heat and can't flex for the latch.

+ dovetail rail cantilever latch spade pockets 14 AWG out FIT=0.2mm tolerance var v3
OpenSCAD · parametric tolerance
FIT = 0.2; // adjust for printer
RAIL_W = 14.5 + FIT;
RAIL_D = 6.0 + FIT;

module latch_arm() {
  difference() {
    cube([40, 12, 4]);
    // hook cutout
    translate([28,-2,0])
      cube([8, 6, 4]);
  }
}
2011 Ford Fusion
Tune & Diagnostics
Automotive · 2024–Present
2.5L I4 · SCT X4 · OBD2

Running a custom tune on a 2.5L I4 via SCT X4 tuner alongside an aftermarket intake and Vibrant resonator + Dynomax Super Turbo cat-back. The car is a rolling diagnostic project — I monitor it the way an engineer would, not just driving it.

Caught and diagnosed an O2 sensor heat soak event at sustained WOT near 110mph — AFR spiked to 21–23 on the datalog, which looked dangerously lean. Traced it to the upstream wideband sensor saturating thermally and reporting garbage, not actual lean mixture (the ECU runs open-loop rich at WOT regardless). The sensor recovered slowly on cooldown, confirming the cause.

Active PIDs being logged: measured AFR, STFT, LTFT, knock retard, RPM, MAF voltage, throttle position. Idle/cruise sits at 14.3–15 AFR — slightly lean-of-stoich, consistent with a minor vacuum leak or MAF calibration offset from the intake swap.

Live datalog snapshot — idle/cruise
AFR (idle)14.3 – 15.0  (slightly lean)
AFR (WOT)12.5  (target rich, open loop)
AFR (fault)21–23  (sensor heat soak)
STFTMonitoring for vacuum leak signature
LTFTLong-term correction baseline
ModsAftermarket intake, Vibrant res, Dynomax muffler
TunerSCT X4 — custom 93-octane tune
AIR / FUEL RATIO RANGE LEAN STOICH RICH 12.0 14.7 17.0 idle 14.8 WOT 12.5 FAULT 21+ (sensor bad)
Wi-Fi Hotspot
VPN Passthrough Applet
Software · 2026
Python · PowerShell · pystray

Built a lightweight Python system-tray app that manages a USB Wi-Fi dongle as a rebroadcast hotspot — taking the laptop's primary connection and sharing it as a separate network to other devices. The UI lives entirely in the system tray with right-click controls, no window.

The key engineering challenge was VPN passthrough — to ensure hotspot clients get tunneled through ProtonVPN or Cloudflare WARP, Internet Connection Sharing must be configured on the VPN virtual adapter, not the raw Wi-Fi interface. Getting this wrong means clients bypass the tunnel entirely.

The app fires netsh wlan and netsh interface PowerShell subprocess calls under the hood, checks adapter state, and presents clean on/off toggle controls. Compiled to a standalone .exe with PyInstaller --noconsole so it runs silently in the background with no console window.

Python · hotspot start via subprocess
import subprocess

def start_hotspot(ssid, password):
    subprocess.run([
        "netsh", "wlan", "set",
        "hostednetwork",
        f"mode=allow",
        f"ssid={ssid}",
        f"key={password}"
    ], check=True)
    subprocess.run([
        "netsh", "wlan", "start",
        "hostednetwork"
    ], check=True)
    # ICS shares from VPN adapter,
    # not raw WiFi — critical for tunnel
Stack
LanguagePython 3
Tray UIpystray + Pillow
Networknetsh (PowerShell subprocess)
VPNProtonVPN / Cloudflare WARP
ICS targetVPN virtual adapter (not raw WiFi)
DistributionPyInstaller --noconsole → .exe
Chromebook CCD Unlock
+ Windows Install
Hardware Mod · 2026
Linux · Suzy-Q · CR50 GSC

Unlocked firmware write protection on a Dell Chromebook 3100 — a modern device with a CR50/Ti50 GSC security chip that controls WP in firmware, not a physical screw or jumper like older models. The only path is CCD (Closed Case Debugging) via a Suzy-Q USB-C debug cable.

Connected the Suzy-Q to a Linux laptop, verified the /dev/ttyUSB1 interface appeared, then used minicom to open the CR50 serial console. Ran ccd open — which requires physically pressing the Chromebook power button several times over ~5 minutes as an anti-remote-attack measure — then wp disable to kill write protection.

With WP off, developer mode was enabled and MrChromebox custom firmware was flashed. Followed the Chrultrabook project for Windows driver compatibility on the Celeron N4000 / Intel UHD 600 hardware.

CR50 console · CCD unlock sequence
# Verify Suzy-Q on Linux laptop
ls /dev/ttyUSB*
# → /dev/ttyUSB0  /dev/ttyUSB1

# Open CR50 serial console
screen /dev/ttyUSB1 9600
# → cr50>

# Check current WP state
cr50> ccd
cr50> wp

# Initiate unlock (press power btn
# on Chromebook ~5x over 5 min)
cr50> ccd open

# Disable write protection
cr50> wp disable
cr50> wp
# → wp: disabled ✓
Hardware
DeviceDell Chromebook 3100 (APRICOT board)
Security chipCR50 / Ti50 GSC
WP methodCCD via Suzy-Q USB-C debug cable
Console/dev/ttyUSB1 @ 9600 baud
FirmwareMrChromebox (mrchromebox.tech)
OSWindows via Chrultrabook drivers
04

Resume

kelton_chelbian_resume  ·  Inferred from conversation history
Kelton Chelbian
I don't wait to be handed answers. I take things apart, figure out how they work, and build something better.
01Who I Am

I'm a high school senior with a hands-on technical mind and a creative drive that pushes me past surface-level answers. I don't approach problems academically — I approach them the way a mechanic or engineer does: What is it actually doing? What's really causing that? How do I fix it?

My interests span automotive diagnostics, electronics and power systems, computer vision and robotics, and software development — and I work across all of them simultaneously. I learn fastest by building something real, breaking it, and debugging my way back.

02Strong Suits
Systems Thinking
I connect cause and effect across complex systems — ECU fuel trims, GPIO conflicts on a Jetson, motor brownout under load. I don't just address symptoms.
Iterative Problem Solving
Test → observe → adjust → repeat. Applied to Li-ion systems, 3D-printed adapter design, FRC vision pipelines, and exhaust decisions.
Resourcefulness
Built high-performance hardware from $15 of harvested Walmart battery cells. Constraints sharpen my thinking rather than stopping it.
Creative Initiative
I don't wait for a defined project. I see something interesting and start exploring it. I generate ideas rather than wait for them to be assigned.
03Technical Skills
AutomotiveECU diagnosticsSCT X4 tuningFuel trims / AFRExhaust fab
ElectronicsLi-ion systemsPWM motor controlBMS / CC-CVArduino
SoftwarePythonJava (WPILib)OpenSCADLinux CLI
RoboticsFRC Team 4156Swerve driveLimelight / AprilTagJetson
CertifiedSACA C-101SACA C-201SACA C-202OSHA
04Real Projects
Robotics● Active
FRC Team 4156 — Vision & Swerve
WPILib Java, Phoenix 6, Limelight 3 AprilTag targeting, NanoOWL on Jetson, full competition season debugging.
Electronics● Completed
DIY High-Performance Water Gun
18650 cell harvesting, OpenSCAD adapter (3 iterations), PWM motor control, BMS integration. Fully original design.
Automotive● Ongoing
Ford Fusion Performance & Diagnostics
SCT X4 tuning, AFR/LTFT/STFT monitoring, exhaust fabrication, O2 sensor diagnostics.
05The Bottom Line

I'm not the candidate who studied the textbook and passed the test. I'm the one who took the engine apart in the driveway, figured out why it was running lean at idle, fixed it, and then went inside to work on a robot. I learn fast, I don't give up when it gets complicated, and I bring genuine curiosity to every problem I touch. That's not something you can teach — it's just how I'm wired.

— Kelton Chelbian, inferred from 50+ conversations with Claude
05

Contact

I'm a high school senior actively looking for opportunities in manufacturing, automation, automotive tech, and robotics. If you're building something interesting, I want to hear about it.