Dillon Schmidt: Fine-tuning BioClinicalBERT on the BC5CDR Dataset

This report covers the work in the Special Projects in AI for ZubAir Data LLC at the SUNY Fredonia CIED Center, completed from mid-January 2026 through May 2026. The program is structured around an applied project and my own work was focused on machine learning and natural language processing, with the goal of getting hands-on experience building and fine-tuning a domain-specific language model end to end.

The main project was fine-tuning BioClinicalBERT on the BC5CDR dataset for biomedical Named Entity Recognition. This domain was selected because healthcare is one of the larger employment sectors in Chautauqua County, so building something with a clinical focus made sense for a project tied to the local area. Using the Hugging Face Trainer API and seqeval for evaluation, I trained the model for four epochs on a single NVIDIA RTX 3070 and ended up with an F1 score of 0.8646 on the test set, which is competitive with published results on the same benchmark.

Krish Chawla: GenAI Prompt Engineering and Chatbot Development

The primary purpose of this project under Special Projects in AI for ZubAir Data LLC at the SUNY Fredonia CIED was to evaluate the practical capabilities of leading generative AI platforms and to apply those findings toward the design and development of deployable AI-powered chatbot solutions for business clients. Core responsibilities included running structured prompt engineering experiments across five major AI tools, researching advanced prompting techniques, investigating security and ethical considerations for enterprise AI deployment, and designing a modular customer service chatbot architecture.