Yewon (Yeni) Ko

Yewon (Yeni) Ko

I'm a mechanical engineering junior at Boston University. I design, build, and program small machines. The current one is a six-legged robot that fits in your hand.

Looking for a Summer 2027 internship in robotics or mechatronics hardware.

Email me

Tripod gait, seen from above A six-legged robot seen from above. Three feet stay on the ground while the other three swing forward, then the two groups trade.
  • Foot on the ground
  • Foot swinging forward
How Tomodachi-kun walks: three feet stay planted while the other three swing forward, then the two sets trade.

About me

I'm a junior studying mechanical engineering at Boston University, graduating in May 2028. I like projects where I get to do the whole loop myself: model the parts in CAD, print and assemble them, wire the electronics, and write the code that makes them move.

So far that has meant a six-legged robot, a candy sorter, a room temperature monitor, and test fixtures and a measurement tool for a materials lab in Korea.

I speak Korean, English, and Japanese. Outside class I'm a member of Theta Tau, the professional engineering fraternity.

What I work with

CAD and design
SolidWorks, Fusion 360, Onshape, nTop
Prototyping and fabrication
FDM and DLP resin 3D printing, laser cutting, manual lathe, thermoforming, soldering and rework
Electronics
Arduino, sensor integration, actuator control, ESC and power-system bring-up
Programming and data
Python, C/C++, MATLAB, Julia, ImageJ, OriginPro
Languages
Korean (native), English (fluent), Japanese (advanced-intermediate)

Projects

Contact

yeni1027@bu.edu

I'm in Boston, MA, and available for an internship from May through August 2027.

Personal project

CAD model of Tomodachi-kun with the top shell removed. Labels point to the ESP32, the PCA9685 servo driver, the six hip servos, a leg, a coxa, the face panel, the shell bosses and the tail block with the start button.
CAD model with the top shell removed.
1 / 1

Tomodachi-kun

Hexapod companion robot

Tomodachi-kun is a six-legged robot about the size of a hand. It is built to walk to a visitor's outstretched hand and to hold the tray on its back level when the surface under it tilts.

Each leg has two MG90S servos: one swings the leg forward and back, the other lifts the foot. An ESP32 drives all 12 through a PCA9685 servo driver.

  • 12servos, two per leg
  • 46%of stall torque on a lift servo while walking
  • 8.4 Aworst-case current, on a 10 A supply
When
Nov 2026 to present
Role
Project lead, team of 2
Built with
SolidWorks, 3D printing, ESP32, PCA9685 servo driver, C++, IMU

What I did

  • Designed the body and legs in SolidWorks, 3D-printed the parts, and assembled the robot.
  • Wrote the C++ firmware on an ESP32 for 12-servo control, the tripod gait, and IMU-based tray leveling.
  • Sized the legs and power system from torque and current calculations.

Sizing calculations

Load on a lift servo while the robot walks on three legs

150 g on each standing leg, 5.5 cm from the lift axis, comes to 0.83 kgf·cm. That is about 46% of the MG90S servo's 1.8 kgf·cm stall torque.

Current if all 12 servos stall at once

8.4 A in the worst case, on a 5 V supply rated for 10 A.

Now
It walks with a tripod gait.
Next
Gait tuning, then time-of-flight sensing so it can walk to a hand.

Course project

Transparent CAD model of the cylindrical color sorter, about three times as tall as it is wide.
CAD model of the full assembly.
Section view of the sorter with a funnel at the top and two servos stacked below it.
Section view with the two servos.
Exploded CAD views of the sorter's printed parts.
Exploded views of the printed parts.
CAD view into the base of the sorter with an Arduino Uno mounted inside.
Arduino Uno mounted in the base.
Dimensioned drawing of the sorter with callouts for the color sensor, wire routing, container, power switch and electronics.
Dimensioned sketch with the main parts called out.
Chart of measured servo angle against commanded angle, a straight line with fit 0.982x + 1.6 and R squared 0.999.
Servo test: measured angle against commanded angle.
Chart of servo response time against load mass, staying between about 218 and 232 milliseconds.
Servo test: response time against load.
1 / 7

Skittles color sorter

Automated color-sorting machine

A small optical sorter that separates Skittles into single-color batches. A color sensor reads each candy, and servo-driven gates route it to the matching bin.

We built it as a team of four. Our accuracy target was 98%, and after tuning we reached 97%.

  • 97%classification accuracy
  • 100+test runs to tune the color thresholds
  • 1 mmor closer: how near the sensor must sit for a reliable reading
When
Spring 2025
Course
EK 210 Engineering Design, Boston University
Team
4 people
Built with
CAD, Arduino Uno (C/C++), TCS34725 color sensor, SG92R servos

What I did

  • Designed the sorting mechanism in CAD and adjusted part tolerances until a continuous feed ran without jams.
  • Wrote the Arduino (C/C++) code that reads the color sensor and drives the servos.
  • Wired and assembled the sensor, microcontroller, and servos.
  • Ran the tests, compiled the data, and tuned the color thresholds across 100+ runs.

What the tests showed

  • Color sensor: the red reading drops off sharply with distance, so detection was only reliable within about 1 mm of the candy.
  • Servo angle: commanded and measured angles matched closely (R² = 0.999), with slight under-rotation (slope 0.982) and a +1.6° offset.
  • Servo load: response time stayed nearly constant as the load increased.

Research internship

Diagram of a droplet on a surface with a circle fitted to its outline and tangent lines at the two contact points marking the contact angle.
What the Python tool automates: fit a circle to the droplet outline and measure the angle where the droplet meets the surface.
Diagram of a fixture held above a test plate at a slant, so the gap between them narrows from 3.5 mm at one end to zero at the other.
The idea behind the tapered-gap fixture: one plasma run covers every gap width from 3.5 mm down to 0.
1 / 2

Plasma treatment study and contact-angle tool

Research internship

A five-week study of how far plasma surface treatment reaches into narrow gaps, done for the lab's work on 3D-printed lattice microneedles.

I designed the test fixtures, ran the study, and wrote the tool that measures the results. Photos and data from the study belong to the lab, so the images here are diagrams.

  • 5 weeksreplicated study
  • 20plates tested
  • 0 to 3.5 mmgap range covered in a single run
When
Summer 2026
Where
Lee Lab (Advanced Additive Manufacturing & Applications), Sungkyunkwan University, Suwon, Korea
Built with
Fusion 360, DLP resin printing, Python, Tkinter

What I did

  • Designed and DLP-printed tapered-gap test fixtures in Fusion 360 that expose a continuous 0 to 3.5 mm gap range in a single plasma run, iterating through vacuum-seal and clamping failures along the way.
  • Ran the replicated study on 20 plates, with seal-integrity controls and reference plates.
  • Wrote a Python tool with a Tkinter GUI that automates contact-angle measurement (droplet detection, circle fitting, tilt correction) and exports results in batches. The lab adopted it.

Course project

A black box with a lit blue LCD showing the temperature, a green power light and a switch on the side.
The finished monitor showing the temperature in °C and °F.
The open enclosure with an Arduino Uno, bundled wiring, a 9 V battery and the LCD mounted in the lid.
Inside: Arduino Uno, soldered and capped wiring, and the 9 V battery.
CAD model of the enclosure with its parts placed inside, and the lid with a cutout for the LCD.
CAD model of the enclosure and lid.
Circuit diagram connecting an Arduino Uno to a temperature sensor, an LCD, two LEDs with resistors, a buzzer, a switch and a 9 V battery.
Circuit diagram.
1 / 4

Room temperature monitor

Arduino monitor with an alarm

A battery-powered monitor that reads the room temperature with a TMP36 sensor and shows it on an LCD. When the room leaves a 70 to 75°F comfort range, a buzzer sounds and a red LED flashes.

I built it start to finish: the circuit, the code, the enclosure layout, and a 3D-printed holder for the 9 V battery.

  • 70 to 75°Fcomfort range before the alarm trips
  • 146 × 120 × 55 mmABS enclosure
  • 9 Vbattery in a 3D-printed holder
When
Spring 2025
Course
EK 131 Hands-On Engineering, Boston University
Role
Individual project
Built with
Arduino Uno (C/C++), TMP36 sensor, I2C LCD, soldering, CAD, 3D printing

What I did

  • Prototyped the circuit on a breadboard, then rebuilt it inside the enclosure with soldered joints, heat shrink, and spade connectors on the switch.
  • Wrote the Arduino code that converts the sensor voltage to °C and °F, updates the LCD, and runs the alarm.
  • Sized the LED resistors with Kirchhoff's voltage law: 1 kΩ for the always-on green LED (about 3 mA) and 220 Ω for the red alarm LED (about 13.6 mA).
  • Modeled the enclosure layout in CAD and 3D-printed a holder for the battery.
  • Checked the readings against a digital thermometer.

Hackathon

A small quadrotor with four black propellers sitting on a wooden table.
The assembled quadrotor.
Hands holding a quadrotor frame with motors and propellers attached, over a workbench.
Fitting the motors to the frame.
A laptop showing a bracket in a CAD program, with a white controller in front of it.
A drone part being modeled in CAD during the event.
1 / 3

One-day quadrotor build

Drone hackathon

Our team set out to build a working quadrotor in one day. It was my first time working on an unmanned aircraft, so I learned alongside the team.

  • 1 dayto assemble and check the quadrotor
When
Feb 2025
Event
HackHardware × AIAA drone hackathon
Role
Team member

What I did

  • Helped assemble the quadrotor and check its power delivery, ESC signals, and frame stiffness.
  • Noted with the team what to improve for stability and flight control.

Club

A scale ride model with metal truss arms over a painted blue base.
The team's ride model.
1 / 1

Ride Engineering Competition

Terrier Ride Engineering Club

Terrier Ride Engineering Club is Boston University's team for the Ride Engineering Competition, where schools design and build a working scale amusement ride. In 2025 I went with the team to the competition at Ohio State, and we placed 5th of 22 schools.

  • 5th of 22schools at the 2025 competition
  • 6+ hoursof continuous running verified by the team
When
Jan to Apr 2025
Team
Terrier Ride Engineering Club, Boston University
Role
Team member

My part

  • I joined in January 2025 as a new member. That season I was mostly learning from the team's design and build work.

What the team built

The team designed the drive and ride structure in SolidWorks, built a low-cost prototype from laser-cut panels and FDM-printed parts, and verified more than six hours of continuous operation under ASTM safety standards.