Joshua Zatz profile pic

Hello I'm

Joshua Zatz

Mechanical Engineer

LinkedIn profile GitHub profile
Loading robot...

Get To Know More

About Me

Profile Picture
Experience Icon

Experience

2+ years
Researcher

Education Icon

Education

B.S Mechanical Engineering @ UIC
M.S Mechanical Engineering | Robotics @ UIC

Hi, I'm Joshua Zatz — a mechanical engineer and robotics enthusiast. I love solving complex problems by combining mechanical design and intelligent software.

I'm currently earning my M.S. at the University of Illinois at Chicago. My research focuses on designing robots that understand and respond to people, a field known as Human‑Robot Interaction (HRI).

Previously I have built sensor systems and computer‑vision pipelines to help robots see and interpret the world in real time. These projects taught me how to bridge hardware and software to create reliable, responsive autonomous platforms.

As robotics advances and computers become more powerful, my goal is to develop machines that safely integrate humans and robots. I see a future where collaborative robots transform how we work, explore, and innovate — and I'm excited to be part of it.

Arrow Icon

Explore My

Experience

Mechanical Engineering

SolidWorks

Experienced

Ansys

Experienced

ROS2 + Gazebo

Intermediate

Machining

Experienced

3D Printing

Experienced

DFM/DFA

Intermediate

Coding

Python

Intermediate

C++

Intermediate

Matlab

Experienced

Git

Experienced

OpenCV

Experienced

TensorFlow

Experienced

View my Resume

▾

Download Resume

Publications

▾

Tuning Squatting Controllers for a Hip Exoskeleton Using EMG-Based Bayesian Optimization

View Publication →

Arrow Icon

Browse My Recent

Projects

FermiLab Computer Vision Robot Arm

▾
FermiLab NUMI diagram
Real flange with beryllium disc
MuJoCo simulation view
0:00 / 0:00
FermiLabs NUMI Project
Goal: Develop an autonomous robotic system capable of detecting and removing bolts from a flange in a radioactive environment. The environment itself is stochastic (lighting, reflections, grime, etc.). The system must be reliable and account for these uncertain conditions.
  • Developed for FermiLab's NuMI project to perform remedial tasks in radioactive environments
  • Integrated UFactory xArm 6 with Intel RealSense Depth Camera for autonomous bolt detection and removal
  • Implemented TensorFlow & YOLOv5 object detection to identify bolts on flanges with identical material and color
  • Overcame challenges of traditional precision techniques through advanced computer vision algorithms
  • Map pointcloud results from LiDAR to the expected flange output from YOLOv5, mapping the appropriate colors (grey-scaled) and bolt locations
  • Use trajectory generation on the array of detect bolts using Inverse Kinematics.
  • Created a robust software for safe, remote operation in hazardous conditions with optional human interference.
Acknowledgements: Completed as part of Senior Design project for FermiLab.
Python TensorFlow YOLOv5 Computer Vision ROS2 OpenCV

Medical Electrolyte Triggered RFID Sensor

▾
Velocity Characterization
0:00 / 0:00
Experimental testing at 10°F
Goal: Design and validate a novel RFID-based medical sensor that activates upon contact with electrolytes, enabling real-time monitoring medicines that can only be outside of a specified temperature range for a specific amount of time.
  • Performed comprehensive experimental validation across multiple temperature conditions
  • Developed detailed multiphase simulation models considering porous microstructure effects
  • Analyzed surface tension variations and their impact on electrolyte behavior
  • Created prototype demonstrating real-time monitoring capabilities in controlled environments
  • Optimized sensor response for medical-grade precision and reliability
Acknowledgements: Completed as part of contract work for Abbott.
Multiphase Flow CFD Simulation ANSYS Fluent Experimental Testing Medical Devices

Supervised Learning on Alloy Hardness

▾
Data extraction
Bulk distribution
Shear distribution
Prediction
TiNi Predictor
0:00 / 0:00
Supervised Learning on Alloy Hardness
Goal: Create a supervised learning model to predict alloy hardness based on chemical properties.
  • Hardness is an empirical emergent property, the best we can do is find correlations to hardness. Density Functional Theory only outputs elastic constants.
  • Important features that show correlation to hardness includes lattice structure and chemical properties (e.g. Bond length is directly proportional to stiffness).
  • Utilized the Materials Project database for input data (100k+ entries) of different alloys.
  • Used Voigt-Reuss-Hill stress to get the mean stress. Ideal for anistropic materials.
  • The correlation and mapping of the composition to moduli/hardness is nonlinear, discontinuous and influenced by environment. For this reason, supervised learning was used (Random Forest Regressor).
  • Random Forest Regressor places importance on certain features, resistant to overfitting, and handles the noisy data well.
  • End results were promising with high coefficient of determination (R²) without over/underfitting.
  • Using the promsing Shear and Bulk Modulus, hardness is calculated via the Chen-Ming Model. Provided as accurate as possible values to hardness of different anistropic alloys.
Acknowledgements: Completed as part of ME494: Machine Learning in Data Science.
Python SciPY/Pandas Supervised Learning Machine Learning

2-Cylinder Inline Engine Fidget Toy

▾
Toy Half Cut
Cam profile
Crankshaft model
Complete Assembly
0:00 / 0:00
Gypsum Quality Refinement & Sustainability
Goal: The task was to design a novel fidget toy that can be easily replicated.
  • Wrote a matlab script that models a singular Cam.
  • Generates the cam outline in xyz coordinates based on Dwell, Rise, and return parameters.
  • Exported the cam profile coordinates onto Solidworks. In series, it created a crankshaft with the desired linear output of the pistons.
  • Performs stress simulation and optimization on crankshaft components.
  • 3D Printed the crankshaft, pistons, and housing.
Acknowledgements: Completed as part of ME347 CAD class.
Matlab SolidWorks ANSYS 3D Printing

Autonomous Stair-Climbing Robot

▾
Assembly view of the Autonomous Stair-Climbing Robot
SW Drawing of robot
Final robot product
0:00 / 0:00
Automatic Stair Climber
Goal: Build an autonomous stair-climbing robot that detects stairs and must climb at least three stairs to qualify.
  • Designed a rack and pinion system fixed at the robot's base. Enables vertical movement of the robot's base.
  • Used button 'antennas' to detect stairs
  • Optimized the motor gearbox to withold the weight of the robot (slower but smoother).
  • Compiled the entire process to automatically detect stairs upon physical interaction, and climb them one stair at a time.
  • Once a stair is in the range of the ultrasonic sensor. The rack and pinion system engages, pulling the robot base up. Extra friction on the wheels allowed for the uneven weight distribution to get pulled. Lightweight back bearings ensured stability while climbing.
  • One of three designs in a class of 250 to actually work and meet the requirements.
Acknowledgements: Completed as part of ME250 class.
Arduino SolidWorks DFM 3D Printing Electronics
Arrow Icon

Get in Touch

Contact Me

Joshzatz@outlook.com