Thomas Jei Kim
Projects
The robots are based on an existing animatronic cat made by Ageless Innovation. We modified the original by adding an embedded custom PCB that includes an ESP32 microcontroller, an MPU6050 accelerometer, as well as a microphone, light sensors and touch sensors located in the cat's eyes, mouth, and body. We designed and built collars that contain an e-ink display and LED strip for user interaction, powered by Raspberry Pi Zero Ws. We included bespoke charging/resting pads for each cat.
ARIES Project
Providence, RI — 2021-2023
The goal of Affordable Robotic Intelligence for Elderly Support is to build robotic cats for older adults. The cats are designed to provide comapansionship for older adults and serve as tools for caregivers. The project is funded by an NSF grant to Professor Bertram Malle's group in the Brown University department of Cognitive, Linguistic, and Psychological Studies.
I was one of two engineering students on the team at any given time, with the project leads all being cognitive scientists. I was the only one with any hardware or robotics experience, so all hardware and embedded systems tasks were my responsibility. Once the lengthy process of setting up the sensors controlled by the ESP32 was finished, the next challenge was to enable communication between the ESP32 in the cats' bodies and the Raspberry Pi in the collars. At first, I was tasked with setting up a wired connection between the chips using the I2C protocol, but after months of trying to make it work, the existing I2C drivers for the ESP32 were just not well-developed enough to be usable. Currently, we are working on setting up a wireless data transfer system. The ESP32 will use its built-in WiFi to upload to a server, and the Rapsberry Pi will download from the server. We intend for the server to also be used by the app we are going to develop.
CLIP-BOSS
Providence, RI — 2023
This project was for my masters seminar"Introduction to Advanced 3D Robot Perception" lead by Srinath Sridhar. We read and discussed the most cutting edge papers in the field of computer vision, which is currently dominated by work involving Neural Radiance Fields (NeRFs).
Our deliverable was CLIP-Based Object Sorting with Spot or CLIP-BOSS. We chose to use OpenAI's CLIP zero-shot object identification program to see if Spot could correctly identify and sort three different objects. Spot's built in locomotion systems made the physical movement of the bot relativley easy, allowing us to focus more on perception. We programmed the robot to put the swirly ball in the basket on its back, the striped ball to one side, and the toy fruit to the other. Integrating the systems from OpenAI and Boston Dynamics was a great challenge in our limited time frame. Combining cutting edge vision software with the newest robotic hardware systems to achieve a high level of functionality left me with great excitement for the future of robotics.
The Robots are Coming!
Providence, RI — 2021
In August 2021, Professor Iris Bahar asked me to help TA a new class she was teaching, called "The Robots are Coming!, The Robots are Coming!". The class focused on the intersection of robotics and art, with help form professional artist Eva Goetz. The sculputres were large wood pieces with train-station style flipboards in the middle.
I designed and built the mechanisms that could run the flipbooks. Based on my design, the students built the remaining eight flipbooks and programmed them to move using Arduino.
I learned the basics of ROS, RViz, MoveIT, amd Gazebo software. On the hardware side, I got to know RealSense depth cameras, and programming over UART. Once I could do basic movements with both machines, I collaborated with the graduate students to apply them to the lab's work. They were working on optimizing Partially Observable Markov Decision Processes (POMDPs) using monte-carlo planning. We chose to use the WidowX250 robot arm because it was easier to program, and not to use of the RealSense cameras to satisfy the "paritally observable" conditions of the study. I 3D printed new grippers for the arm and attached force sensors to them so it could explore a 3x3 grid in a partially observable way.
Bahar Lab
Providence, RI — 2021
During the Summer of 2021, I worked in Professor Iris Bahar's lab in Brown University's School of Engineering. I was tasked with getting two robots working and finding applications for them within the scope of the labs' projects. We had a locobot WX200 rover and a WidowX250 robotic arm. Neither robot had been touched since 2019, and their software was no longer being updated. I had to directly contact the original developers of the drivers and do significant maintenance on my end in order to get them ready for use.
DanceBot
Provience, RI – 2021
This was a final project for the introduction to robotics class in Spring 2021. Our goal was to build a dancing robotic sculpture using only a single arduino, four small servo motors, and any scrap material we could find in the University's workshop. We connected the motors to the central "dancer" and used plastic rods to connect the others and make a unified system. We lasercut the pieces and were able to power everything off a single 9V battery. Many long nights were spent cutting parts, coding, and assembling the structure, but in the end we had a product that made us proud.
Para Inc.
Remote — 2020
Para is a company dedicated to making gig work (e.g. Uber, Doordash, etc.) better for gig workers. The major apps all collect large amounts of data about their workers, but do not provide it to them. Drivers can greatly benefit from knowing their $/hour, $/mile, and which times of day, locations, customers, and apps are most profitable.
I joined the team as a Summer intern in 2020, working remotely during the peak of the COVID-19 Pandemic. When I joined, the engineering team consisted of the co-founder and CTO, myself, and one other intern. The engineering work consisted of running a flask-based website that asynchronously pulled data from the gig apps' APIs and brought it into our databases where it could be analysed and displayed to the users. It was my first time working in the front-end heavy webapp space, and using database products like Postgress as a customer. I also got to participate in weekly all-hands meetings, in which we discussed how we were going to grow the company not just through better engineering, but also through user and investor relations. I especially admired how David Pickerell, the CEO, was driving Uber himself and spending most of his time talking to gig workers and trying to understand what sorts of features they wanted. Talking to investors and getting funding was important, but never more important than doing good by our end users. I found myself pitching our company to every Uber driver I ran into.
Today, Para is more than just a data source for drivers. It serves as a platform for drivers to talk to each other, and with the newly deployed ParaWorks feature, it gets high-paying jobs directly form businesses without even going through the apps.
Snowflake Computing
San Mateo, CA — 2019
When I joined Snowflake for a software engineering internship in Summer 2019, they were in the final stages of the startup life cycle, and preparing for their IPO. It was my first time working on production code, at a silicon valley company, and in a corporate environment. I was lucky enough to be part of a special program working with another young intern directly under the founder and CTO, Thierry Cruanes. I got to see firsthand what it takes to build and sustain a major tech firm.
Our main project was to improve error messages in customer-facing software, and to build infrastructure that would translate the product to global locales. We had to communicate with teams across the company to get descriptive, concise error messages for every possible case. For the locale translation piece, we learned what is was like to bulid a new program from scratch and integrate it into the company's codebase. I learned that the key to enterprise-level software engineering is communication and documentation. Well-written code must be more than functional, it must be reproducible and easy to fix in the inevitable case of bugs. I also learned that even though Snowflake had a massive valuation and was in the news, it did not have the impact on the world that I wanted to have in my work. Snowflake's users are mostly data-engineers at major corporations, and their long-term goal is fundamentally greater corporate growth. I left with the impression that I wanted to explore work in other areas.
Today, Snowflake is a world leader in data warehousing services. They provide data management and analytics for any business that keeps their data on Amazon's S3, Google's GCP, or Microsoft's Azure servers. They've been on Forbes's Cloud 100 list, and IPO'd at a valuation of $3.4 billion. Their customers include AT&T, Capital One, JetBlue, and DoorDash.
Brown Sports Conversations
Providence, RI — 2018-2022
The Brown Sports Conversations club was founded in 2018 as a place where die-hard, casual, and newcomer fans could come together to enjoy the culture and spectacle of all types of sports.
Our activities included organizing get-togethers like playing Around the Horn, throwing watch parties, and of course lots of banter. We also started a website to post articles and the Brown Sports Convos Podcast available on Spotify, Apple music and Soundcloud.
I was a founding member of the podcasting team alonside my podmates David Coady, Samuel Rhee, and Ethan Jobson. I learned how to speak clearly yet passionately, and be comfortable in front of a microphone. We navigated the challenges of setting up and using recording equipment in a studio, and later doing pods over zoom.
We interviewed several Brown University athletes, including Shayna Mehta from Women's Basketball, and Abby Carchio from the back-to-back-to-back Ivy League Champion Women's Soccer team. The interviews presented a new set of challenges. We had to prepare questions that kept our guests engaged and eager to share, while also being exciting content for our listeners.
Though at first it felt like we were recording only for our friends, by senior year, I was getting compliments from students I had never met who were listeners.