Hi, I'm Connor Pennington
Software & AI Engineer
Passionate about creating innovative solutions with machine learning and artificial intelligence. Currently studying Computer Science at Clemson University.
Experience
NVIDIA
- Benchmark agentic workflow frameworks (AWorld and AgentOrchestra) on the GAIA dataset, where state of the art models are replaced with a Small Language Model (SLM).
- Create a dataset for finetuning the SLM from agent failure patterns, traces, and distilled data focused on Metropolis use cases, such as public safety, warehouses, and traffic.
- Finetune and evaluate the model, measuring against the baseline out-of-the-box SLM.
Sixfold AI
- Improved the PDF data extraction pipeline in Python, increasing system accuracy by 8%.
- Integrated an open-source LLM chat interface with a Model Context Protocol (MCP) gateway and multiple servers, enabling workflow automation that saved employees 10+ hours weekly.
- Collaborated closely across teams of different roles to define success metrics and applied LLM evaluation methods (pairwise comparisons and accuracy) to validate and refine automation.
OpenAI
- Built a prompt suite to test and guide the moral reasoning capabilities of LLMs.
- Implemented a system where a less advanced model generates complex moral scenarios for a more advanced model to analyze and respond to based on the moral implications presented..
- Designed evaluations to analyze moral learning effectiveness across 3 different model gaps.
402 SOFTWARE ENGINEERING GROUP
- Utilized a machine learning model to extract tables from over 600 uniquely formatted PDFs.
- Engineered a C# pipeline to convert the extracted tables from CSV to XML files, eliminating manual table processing and boosting development productivity by approximately 20%.
Research
Humans and Technology Laboratory (HATLab)
- Design prompts for LLMs to create inferences and recommendations on participant data.
- Build a pipeline that inputs raw data and outputs a Sankey Diagram of inferences and recommendations.
- Implement a RAG system to match LLM recommendations with real products from Amazon.
Big Data Analytics Lab
- Develop a web scraping tool to download open-source material science research articles.
- Parse and store text and image elements from documents for Multimodal RAG.
- Create a chat-bot application that allows users to interact with the RAG enhanced LLM.
Advanced Imaging Lab
- Benchmarked state of the art models for 3D scene reconstruction.
- Submitted paper to WACV 2025 as co-author.
Sensors, Materials, and Analytics for Regenerating Tissues (SMART) Lab
- Redesigned Python scripts to reformat amino acid data.
- Developed a user interface, allowin users to easily input their amino acid data to train neural network models.
Featured Projects
Lecture Genie
An intelligent lecture analysis tool that processes educational videos using AI. Features automated transcription, content summarization, and study tool generation to enhance learning experiences.
Climate Crisis Film Repository
An engaging platform for exploring the climate crisis through film, featuring a curated collection of climate-focused movies and expert analyses for deeper understanding. It also offers personalized film recommendations and discussion boards to connect with others passionate about the environment.
Skills & Technologies
Programming Languages
- Python
- C++
- C
- SQL
- Java
- Ruby
- JavaScript
- C#
AI & Machine Learning
- TensorFlow
- PyTorch
- Agent Frameworks
- Model Context Protocol
- RAG
- Finetuning
- NeRF
Tools & Technologies
- Git/GitHub
- Docker
- Linux
- Postman
- Cluster Computing
- Cursor
- AWS
Education
Clemson University
- GPA: 4.00/4.00
- Clubs: AI Club, Latin Dance Club, Book Club, Intramural Ultimate Frisbee, and Clay Club
- Study Abroad: Cryptocurrency and Blockchain Experience in Nicosia, Cyprus.
Let's Connect
I'm always interested in new opportunities and exciting projects. Let's build something amazing together!