Adam Williams

Robotics Engineer, Software Developer, Data Analyst, Machine Learning Hobbyist and more

Education

MSc in Mechanical Engineering

Virginia Tech

2016 - 2018

BSc in Mechanical Engineering

Duke University

2009 - 2013

Interests

Basketball
Hiking
Photography
Cooking with my wife

About Me

After working for three years as a field commissioning engineer with TMEIC, I decided to return to school to get my M.S. As a researcher at the Robotics and Mechatronics Lab at Virginia Tech, I had the opportunity to gain valuable knowledge and experience working with robotics, mechatronics, controls, and design.

My wife and I moved to Southern California, where I have had the opportunity to work on some fascinating technical challenges at both Adsys Controls and Kulicke & Soffa. At Adsys, I worked as a control engineer on high enery laser systems and designing computer vision algorithms. I am currently working at Kulicke & Soffa in their ultrasonic group, using Python to automate our testing and performing analysis on next-gen ultasonic transducer designs.

I love the challenge of being faced with a difficult problem and coming up with a clever and creative engineering solution, from my field engineering days to my research experience. There is no better feeling than seeing your efforts result in a process line starting up again, or a robot starting to move or a pattern begin to emerge from your analysis.

Experiences

Senior Mechanical Engineer

2019 - Present

Kulicke and Soffa, Santa Ana, CA

In the ultrasonic group at Kulicke and Soffa, I lead the development of Python software used to measure, characterize, and analyze ultrasonic transducer performance in an automated test. My tools are used at our production site, and I have leveraged the data collected by my software to perform statistical analysis on new designs. I have presented my analyses on transducers to internal upper level management as well as for critical customers' engineering decision makers. I have also developed algorithms for automatic paramter tuning used in our ultrasonic generators, and designed a novel sensor that can be used to more accurately measure the performance of an ultrasonic transducer.

Staff Engineer

2018 - 2019

Adsys Controls, Irvine, CA

While at Adsys Controls, I was the lead controls engineer for several interesting projects. I worked to build a state space model of a complex optical system in Simulink, while on the side built a custom ray tracing program in order to design a non-axisymmetric optic. After that, I led the development of an algorithm for detecting and tracking drones from a moving platform, incorporating inertial feedback as well as optical input. That project was the first time I used Python in a professional sense, and was a great crash course in developing reusable, human-readable, efficient code.

Graduate Research Assistant

2016 - 2018

Robotics and Mechatronics Lab, Virgina Tech, Blacksburg

As a graduate research assistant, I was responsible for conducting rigorous robotics research with the ultimate goal of publishing my work in peer-reviewed conferences and journals. This research was conducted both individually as well as in teams. Personally, I felt I was most successful working in large and small teams, as I enjoy talking through ideas because not only do team members have excellent suggestions and strategies, but also talking out an idea helps me think it through more clearly myself!

Field Commissioning Engineer

2013 - 2016

TMEIC, Roanoke

My role as a field engineer was both to set up, troubleshoot, and perform application-specific tuning for medium-voltage hardware as well as to diagnose and repair problems in already installed systems. After working for my first 6 months in Level I steel mill automation, I transitioned to working on variable frequency drives (VFDs). I spent time all over the world commissioning drives for all kinds of motor control applications. While it was a high-pressure job, especially showing up to a site when the manager immediately informing you the plant is losing tens of thousands dollars during the outage, the feeling after fixing the problem was extremely rewarding.

Pratt Undergraduate Research Fellow

2012 - 2013

Microscale Physicochemical Hydrodynamics Laboratory, Duke University, Durham

After being awarded a Pratt Undergraduate Research Fellowhship, I performed materials science and hydrodynamics research at uPhyl. During my time in the lab, I investigated the action of water droplets on the wing of the lacewing insect. I discovered a novel phenomenon in the manner in which droplet coalesced and "jumped" from the hairy fibers covering the wing, and replicated it with synthetic fibers. The work eventually resulted in a publication in Physical Review Letters.

Publications

K. Zhang, F. Liu, A. Williams., X. Qu, J. Feng, C.H. Chen, Self-Propelled Droplet Removal from Hydrophobic Fiber-Based Coalescers Physical Review Letters, Vol. 115, 14 August, 2015. Reported in Physics

A. Williams, W. Saab, P. Ben-Tzvi, Analysis of Differential Mechanisms for a Robotic Head Stabilization System , Proceedings of the 2017 ASME IDETC/CIE, 41st Mechanisms & Robotics Conference, Cleveland, Ohio, Aug. 6-9, 2017.

B. Sebastian, A. Williams, P. Ben-Tzvi, "Control of a Head Stabilization System for Use in Robotic Disaster Response" , Proceedings of the 2017 ASME International Mechanical Engineering Congress and Exposition (IMECE 2017), Tampa, Florida, Nov. 3-9, 2017.

B. J. B. Lee, A. Williams, & P. Ben-Tzvi(2018). Intelligent Object Grasping With Sensor Fusion for Rehabilitation and Assistive Applications. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 26(8), 1556–1565.

V. R. Kamidi, A. Williams, & P. Ben-Tzvi (2018). A Framework for Modeling Closed Kinematic Chains with a Focus on Legged Robots. 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2733–2738.

B. Sebastian, A. Williams, & P. Ben-Tzvi (2018). Gaussian Kernel Controller For Path Tracking in Mobile Robots. International Design Engineering Technical Conferences and Computers and Information in Engineering Conference 2018 Aug 26

A. Williams, B. Sebastian, & P. Ben-Tzvi (2019). A Robotic Head Stabilization Device for Medical Transport. Robotics, 8(1), 23.

A. Williams, B. Sebastian, & P. Ben-Tzvi (2019). Review and Analysis of Search, Extraction, Evacuation, and Medical Field Treatment Robots. Journal of Intelligent and Robotic Systems: Theory and Applications, 96(3–4), 401–418.

Projects

Here are a few of the projects I have worked on in the lab, at home , or in my career

March Madness Machine Learning Model - Since the 2018 tournament, I have been creating a machine learning model in order to predict the outcome of the NCAA basketball tournament. I was originally inspired by the Kaggle March Madness competition, and successfully have applied my model to win the pool with my college friends for 3 of the last 4 years (My best finish in the Kaggle competition was top 16%, but the bragging rights and $$ in the pool are much more valuable). I have experimented with a variety of models, from logistic regression, random trees, SVM, and various configurations of MLP. I enjoy trying new models every year, and also new machine learning packages. I initally used Keras/Tensorflow and more recenetly moced to PyTorch. I included a loss plot I generated from the PyTorch model during a quick training run, plus the results of a quick run. Picking 63% of the games correctly isn't too shabby!

NBA Machine Learning Applications - As a project in my spare time (which I was able to turn into a class project as well!) machine learning techniques were implemented to analyze NBA data. Principle component analysis (PCA) is utilized to reduce dimensionality of player statistics, which are then aggregated by team and utilized to train a support vector machine (SVM) to predict winners and losers of individual games. K-means clustering is used to detect players whose contributions to their team's success are not adequately represented by statistics. In the first image below, the results of PCA on the statiscal contributions of each player in the league are shown. The clear grouping of players led to the analysis in the next image. This one shows the groups resulting from k-means clustering, including groups 3 (All-Star do-it-all NBA forwards) and 4 (Sharpshooting All-Star guards). Finally, the last image shows the a cross-section of the decision boundary for utilizing teams combined PCA component vlaues from the previous year to predict winners in the following year.

Differential Drive Robot Control - Another project that I played with was the modeling of a discrete differential drive robot (to help out the brilliant Dr. Bijo Sebastian in some of his research) in Matlab first then eventually Python. The goal was to create a basic simulation for tracked robotic control while following a path, so additionally I also modeled the discrete form of a couple controllers. The images below show the results of the simulation for a PID controller.

Robotic Head Stabilization System In order to keep a patient’s cervical spine stable during transport on the casualty rescue robot I worked on during my M.S., an active stabilization device was designed and built. Utilizing active disturbance rejection control, the system maintains the stability of the patient’s head with a single actuator through a compliant mechanism. Below is a diagram of the proof-of-concept prototype from my IDETC conference paper explaining the original mechanism design process, along with a little GIF showing the operation of the physical prototype. A conference paper on early control simulations performed on the inital design can be found here. A totally redesigned final prototype with a more compact form factor, sleeker design, and implementing the active disturbance rejection control scheme is the subject of a journal paper here

Assistive Robotic Hand Orthosis - Improved upon existing robotic hand exoskeleton designs built at RML in order to design a low-profile orthosis that would provide intelligent grasping. The orthosis implemented a slip-aware grasping paradigm in order to enable intuitive operation, which is the subject of an upcoming journal paper. Pictured is the original exoskeleton system that was modified in our work, published in the original paper here for those that are interested.