About
Hello! I'm Jordan, a University of Washington student and researcher, a budding ML engineer, and a guy slighty obsessed with code quality. I’m currently looking for Computer Vision or Machine Learning related work.
Download my ResumeMy Skills
Python
C
R
JavaScript
Java
Verilog
What i do
Machine Learning Engineering
As a Machine Learning Engineer, I'm fascinated by Computer Vision and have experience with training and using supervised and unsupervised machine learning models for signal processing applications.
Software Development Engineering
As a Software Developer, I'm experienced in leveraging data structures and algorithms to effectively address challenges and deliver refined solutions, while minimizing the runtime and memory usage.
My recent Works
0
Years Experience
0
Projects Delivered
0
Seconds on this site!
Headlights
I have developed an innovative text-based video game that leverages advanced natural language processing (NLP) techniques to understand and interpret free-form English commands given by players. This project showcases my comprehensive skills in programming, machine learning, and software development, as well as my ability to work with cutting-edge technologies and adapt them to create engaging user experiences.
During the development process, I utilized a combination of Prolog and Python programming languages to create a robust and efficient game engine. My choice of Prolog allowed me to harness its unique strengths in symbolic reasoning and knowledge representation, while Python enabled me to implement various machine learning models and NLP algorithms efficiently.
To ensure the seamless deployment and scalability of the game, I leveraged Amazon Web Services' Elastic Container Service (ECS) as the underlying infrastructure. This allowed me to efficiently manage and scale the game's resources as needed, providing a consistent and enjoyable gaming experience for players.
A significant aspect of the game's success is its reliance on the state-of-the-art NLP engine, Perplexity. I had the opportunity to contribute to the development of Perplexity during my internship at Inductor Software, where I gained hands-on experience in implementing and optimizing NLP algorithms. Additionally, I wrote regression tests for the engine, ensuring its reliability and performance.
To further enhance the game's NLP capabilities, I utilized Cyberduck to collect user data, which enabled me to analyze player interactions and refine the game's understanding of natural language commands. This iterative process of improvement allowed me to deliver a more immersive and intuitive gaming experience that adapts to individual players' language preferences and styles.
In summary, my advanced text-based video game showcases my expertise in natural language processing, programming, and cloud-based infrastructure management. As a result, I have demonstrated my ability to create compelling and technologically advanced products that cater to users' needs and preferences. My commitment to continuous improvement, adaptability, and understanding of cutting-edge technologies make me a valuable candidate for any team seeking a talented software engineer or machine learning expert.
PCA, Clustering
In this project, I successfully implemented and applied advanced machine learning techniques such as Principal Component Analysis (PCA) and Clustering (k-Means and Gaussian Mixture Model) to analyze and process high-dimensional data. This experience has significantly contributed to my growth as a data scientist, and allowed me to develop and showcase my proficiency in using Python for machine learning tasks.
Throughout the project, I demonstrated a deep understanding of the underlying mathematical concepts and computational methods involved in PCA and Clustering. I implemented PCA from scratch using the NumPy library, which enabled me to effectively reduce the dimensionality of the ORL Face Data dataset while preserving as much information as possible. This hands-on experience has enhanced my understanding of dimensionality reduction techniques and their applications in real-world scenarios.
In addition to PCA, I implemented the k-Means clustering algorithm from scratch and utilized the Gaussian Mixture Model (GMM) clustering algorithm from the scikit-learn package. I employed these clustering techniques to analyze and classify the well-known MNIST database of handwritten digits. This part of the project required me to make informed decisions on the number of clusters and the optimal dimensions for PCA reduction, which has further developed my critical thinking and problem-solving skills.
Throughout the project, I showcased my strong Python programming skills and my ability to leverage popular libraries such as NumPy, pandas, matplotlib, and scikit-learn. I also demonstrated the importance of clear and effective data visualization when presenting complex results, which is essential for conveying meaningful insights to both technical and non-technical audiences.
Overall, this project has allowed me to further develop and demonstrate my expertise in machine learning, data analysis, and programming. I am confident that my experience with this project will serve as a strong foundation for future endeavors in the field of data science and machine learning.
Image Processor
As a part of my academic pursuits in EE 440 at the University of Washington, I developed an image processing tool using Python, Tkinter, and OpenCV. The project provided me with invaluable insights into the practical applications of image processing techniques, and it allowed me to hone my skills in software development and user interface design. The tool I created is capable of performing various image processing operations such as blurring, saturation adjustment, contrast adjustment, temperature adjustment, adding noise, and exposure adjustment. By creating this tool, I gained a deeper understanding of the underlying algorithms and techniques used in image processing and further developed my proficiency in Python and its associated libraries. This project has been an important milestone in my professional development, showcasing my ability to design and implement complex software solutions.
Multi-Object Tracking System
Within this project, I have developed an exceptional solution that combines advanced techniques to revolutionize multi-object tracking. With a focus on accuracy, scalability, and real-time performance, this system stands as a testament to my expertise in computer vision and machine learning.
By seamlessly integrating robust scaling preprocessing, feature matrix integration, and state-of-the-art deep learning-based object detection models, I have achieved unprecedented levels of precision in detecting and tracking multiple objects in video sequences. The system's ability to handle complex scenarios, such as occlusions and scale variations, showcases my ability to tackle real-world challenges with ease.
Moreover, I have taken great care to ensure the system's practicality by incorporating predictive capabilities for future object location estimation. This not only enhances tracking accuracy but also underscores my commitment to providing innovative and forward-thinking solutions.
With a keen eye on industry needs, I have integrated popular deep learning-based object detection models into the system, showcasing my ability to leverage the latest advancements in the field. The result is a seamless fusion of cutting-edge technologies, reinforcing my position as a valuable asset in developing state-of-the-art solutions.
The Multi-Object Tracking System's documentation reflects my attention to detail, making it easy for ML engineers of various backgrounds to integrate and utilize the system effortlessly. As you explore the repository, you will find comprehensive instructions that empower users to leverage its capabilities efficiently, further highlighting my commitment to delivering practical solutions without unnecessary complexity.
Multi-Modal Image Analysis
I have had the privilege of working on an exciting project centered around Multi-Modal Image Analysis, a cutting-edge field within computer vision. Throughout this endeavor, I tackled several significant challenges in the domain, showcasing my expertise and passion for advancing the field. One of the primary objectives involved the development of a robust Convolutional Neural Network (CNN) model using the CIFAR-10 dataset. By skillfully leveraging deep learning techniques such as convolutional and fully connected layers, I successfully achieved exceptional accuracy on the CIFAR-10 test set, demonstrating my ability to optimize model architecture and parameters effectively.
In addition, I delved into the complex realm of imbalanced data and long-tailed recognition, a critical issue in image analysis. By working extensively with the CIFAR-100 dataset, I constructed an imbalanced variant called CIFAR-30, strategically manipulating class distributions. Employing resampling and reweighting techniques, I made remarkable strides in improving model performance, especially for tail classes. Furthermore, I conducted a comprehensive comparison between the imbalanced CIFAR-30 dataset and a balanced counterpart, further solidifying my expertise in handling real-world challenges and delivering impactful solutions.
Lastly, I delved into the captivating world of Generative Adversarial Networks (GANs) by implementing one using the CelebA dataset. The objective was to generate realistic and diverse facial images of celebrities. By skillfully training a generator and discriminator network in an adversarial setup, I successfully showcased the immense potential of GANs in image synthesis. This accomplishment further underscores my ability to harness deep learning techniques and apply them to create novel and visually impressive outputs.
Overall, this project represents a testament to my dedication and expertise in multi-modal image analysis. From building powerful CNN models for accurate image classification, addressing imbalanced data challenges, to demonstrating proficiency in GANs for image generation, I have consistently pushed the boundaries of what is possible in computer vision. I am confident that my contributions to this project exemplify my ability to tackle complex real-world problems and make significant advancements in the field.
Semantic Vision
I have engineered an ambitious personal project called Semantic Vision, which focuses on advancing the field of computer vision by tackling challenging tasks such as object detection and semantic segmentation. Through the utilization of cutting-edge techniques and state-of-the-art frameworks like Detectron2, I have delved into the intricacies of object detection, including fine-tuning pre-trained models and conducting meticulous evaluations. Moreover, I have successfully developed a sophisticated Convolutional Neural Network architecture capable of accurately segmenting diverse images into various classes, such as balcony, window, pillar, and facade, showcasing impressive results on the Mini Facade dataset. By meticulously documenting the detailed implementations, model architectures, training configurations, and evaluation procedures in Jupyter Notebooks and code files, I have demonstrated my proficiency in both practical implementation and theoretical understanding of computer vision algorithms.
Through the Semantic-Vision project, I have demonstrated my prowess in the realm of computer vision, showcasing a keen ability to leverage state-of-the-art frameworks and algorithms to solve complex problems. By meticulously fine-tuning pre-trained models and training custom networks, I have achieved remarkable accuracy in object detection tasks, allowing for precise identification and localization of objects within diverse datasets. Moreover, my development of an advanced Convolutional Neural Network architecture has revolutionized the process of semantic segmentation, enabling precise class-based segmentation of images with exceptional results on the Mini Facade dataset. My rigorous documentation of the project's implementations and methodologies further exemplifies my dedication to clarity, reproducibility, and effective communication of complex ideas.
This project stands as a testament to my expertise in computer vision and serves as a strong indication of my capabilities as a machine learning professional. With an emphasis on both theoretical understanding and practical implementation, I have successfully addressed challenging problems in object detection and semantic segmentation. My utilization of state-of-the-art frameworks, meticulous evaluations, and the development of advanced network architectures all contribute to a comprehensive showcase of my skills and knowledge in the field. This project not only highlights my ability to tackle complex tasks but also underscores my commitment to pushing the boundaries of computer vision research and development.
Stock Price Predictor
I have spearheaded the development of a sophisticated stock price prediction tool, integrating cutting-edge machine learning techniques with a user-friendly graphical interface. This project leverages Long Short-Term Memory (LSTM) neural networks to analyze historical stock data and forecast future stock prices. By meticulously designing and fine-tuning the model's architecture within TensorFlow/Keras, I have optimized the prediction accuracy, enabling precise daily closing price forecasts. My work demonstrates a deep understanding of sequence modeling and its applications in financial data, with the flexibility to expand the tool’s capabilities for diverse market scenarios.
Through this project, I have showcased my ability to bridge the gap between advanced machine learning algorithms and practical user applications by developing a fully interactive graphical user interface (GUI) using PyQt5. The application empowers users to easily customize key prediction parameters, such as the training time frame and number of epochs, while providing real-time feedback on training progress and prediction performance. My extensive work on building intuitive user interfaces and seamlessly integrating machine learning models exemplifies my dedication to creating accessible and impactful data-driven tools.
This project is a testament to my proficiency in machine learning, particularly in applying LSTM networks to time-series data, and my expertise in developing end-to-end software solutions. By integrating financial data retrieval from the Alpha Vantage API, implementing a dynamic LSTM model for prediction, and developing a visually compelling data visualization framework with Matplotlib, I have demonstrated my ability to solve complex, real-world problems. This project not only underscores my technical skills but also highlights my commitment to creating functional, reliable, and user-centric applications in the rapidly evolving field of financial technology.
Check-In
Through my Check-In project, I developed a machine learning solution designed to estimate room cleanliness and mental health suitability using advanced computer vision techniques. Leveraging a pre-trained ResNet50 model on ImageNet, I created a robust data preprocessing pipeline that includes resizing, normalization, and standardization. Additionally, I implemented a comprehensive data augmentation strategy, employing techniques such as rotation, cropping, brightness adjustment, and color jittering to simulate diverse environments. This rigorous approach has resulted in a high-performing classification model capable of assessing environmental factors that contribute to psychological well-being, achieving an impressive 96.59% accuracy across test data. The success of this project showcases my expertise in computer vision, deep learning, and data preprocessing, along with my ability to apply these techniques to real-world problems.
SIGHT
In the SIGHT project, I played a pivotal role in developing a cutting-edge solution that leverages drone imaging and machine learning to detect humans in wilderness environments for search and rescue operations. This project utilized advanced image processing techniques alongside a deep learning pipeline based on the YOLO (You Only Look Once) object detection model. We trained our model on thousands of annotated drone images, collected under various wilderness conditions, to improve its accuracy in detecting humans. Despite challenges such as occlusion and varying environmental factors, we achieved a model precision of 63% with an identification time of 105 ms, showcasing our ability to balance performance and computational efficiency in real-world scenarios. This project highlights my expertise in computer vision, deep learning, and data collection under challenging conditions, demonstrating my ability to apply machine learning models to critical real-world problems with life-saving implications.
Panorama Image Stitcher Using OpenCV
I developed a deep learning model to classify malaria-infected and healthy blood cells using the Malaria dataset from TensorFlow. The project involved building a Convolutional Neural Network (CNN) from scratch and progressively improving its performance through techniques like data augmentation, transfer learning with MobileNetV2, and regularization using dropout layers. The model achieved high accuracy and an F1 score, making it effective for distinguishing between infected and uninfected cells. Additionally, I visualized the model’s predictions to better understand its performance in real-world scenarios.
Panorama Image Stitcher Using OpenCV
Working on this project allowed me to dive into the fascinating world of image processing and computer vision. This tool takes multiple images with overlapping areas and merges them into seamless, high-resolution panoramas, perfect for photographers, developers, and hobbyists!