Bradley Peterson

Accelerated Computer Science graduate student at ASU (B.S. '24, M.S. '25)

Specializing in NLP, LLMs, and ML engineering / research

Astrophotographer and DarkSky International Delegate

NLP Toxicity Mitigation Research Project

Following my CS senior capstone project, I've continued an investigattion into the potential for fine-tuned LLMs to promote a more constructive peer review environment, in collaboration with peers and professional researchers. To help achieve our goals I developed an automated technique for augmenting our value-aligned training data. Additionally, I've utilized my machine learning skills to expand our dataset further, guide our progress measurement, and aid in fine-tuning three open source LLMs to successfuly identify and rephrase toxic and concerning sentiments.

Local Light Pollution Analysis

Motivated by my deep appreciation for night skies, this project combines astrophotography with data science to help combat light pollution. I have collected and processed 8,900+ zenith sky brightness samples across central Arizona, writing ImageJ scripts and utilizing exiftool for image data extraction. I've used tools and libraries such as scikit-learn, k-means, DBSCAN, and Keras to classify and compare lighting strategies from various parts of The Valley and predict sky brightness from urban, atmospheric, and temporal features. Through analysis and visualization using Python and QGIS, I delivered actionable recommendations to local city councils, influencing city lighting objectives. This project not only showcases my technical skills and adaptability but also underlines my passion for leveraging technology for meaningful causes.

AI Travel Assistant - Powered by GPT-4

As an extracurricular collaboration with two W.P. Carey Business School students at ASU, I created a conversational interface for what can sometimes be a tedious or intimidating process: travel planning. Powered by the emergent reasoning capabilities of GPT-4, this app functions as a powerful travel assistant, capable of using its vast knowledge in combination with a variety of travel APIs in order to search and book flights, hotels, vehicle rentals, and a variety of activities and points of interest, as well as comprehensively plan trips end-to-end.

p5.js Projects: Games, Simulators

I've built a wide variety of projects with JavaScript, particularly p5, a dynamic animation library. I started my CS journey creating a few simple games, calculators, and simulators, eventually building an arcade-style fixed shooter and a quadtree-optimized interactive physics simulator, highlighting my skill with arrays and data structures. Later, in collaboration with a group of fellow Ira Fulton Engineering students at ASU, I created a series of games as part of a class project. Thanks to my experience with p5, I took the lead on this project, directing our project's structure, and personally developing the more challenging games we decided to build. Our project was a strong success, earning our group an A+.

Skills

A curated overview of my technical and interpersonal skills. From hands-on software development in multiple languages to effective team leadership and adaptive learning, my skill set is both diverse and specialized, crafted over years of projects, collaborations, and continuous learning.

Programming

I am very experienced in Python, having developed a variety of small applications, utility scripts, and algorithmic solutions, including work on efficiency and safety metrics in the automotive industry. I've developed interactive games with JavaScript and various Java and C++ programs for both personal and academic purposes.

More Detail

Data Analysis

I recently completed a 6-month internship in which I found and presented insights on a wide array of data on a daily basis. I'm also conducting an ongoing light pollution research project, processing and analyzing images to identify both local and long-term urban and temporal phenomena using regression and clustering. My findings have influenced local lighting policies.

More Detail

Communication

My most recent internship involved frequent interaction with individuals in multiple departments across the company, and expecially within my own team. In my previous position at OFW, I acted as a liaison between Data and other departments, translating technical issues into actionable solutions. Additionally, I regularly present on dark-sky advocacy to a variety of audiences, highlighting my presentation skills.

More Detail

Prompt Engineering

I have integrated Large Language Models like ChatGPT into applications, balancing model trade-offs and optimal prompting techniques, and have developed an automated annotation strategy to accelerate the collection of reliable and value-aligned natural language data.

More Detail

Web Development

I have some experience with CSS, HTML, and React, having developed the front-end for my Conversational Travel Planner App and modified/expanded the features of this site. Always keen to expand my front-end skills.

More Detail

Machine Learning

I have experience with building a variety of models and datasets both in school and independently. In an ongoing project, I'm applying my knowledge of various models and ML techniques to predict and interpolate sky quality spatiotemporally based on several dozen urban, weather, and time-based features.

More Detail

Leadership & Teamwork

Over the course of my recent internship, I interacted frequently with several groups and individuals within the Software department and worked extensively with a Data Team of five. I have also led several group projects at ASU, often driving technical direction while ensuring strong participation. I collaborated with business students on an AI Travel Assistant App, bridging technical and business perspectives in the development of a unique and functional prototype.

More Detail

Adaptability & Learning

I continuously adapt to new technologies and challenges, most recently delving into the world of Large Language Models. I am passionately motivated by opportunities to learn and evolve.

More Detail