Certificate of Completion

This certifies that

Jingwen Wang

has successfully completed the hands-on workshop

Operation Codenames: Fine-Tuning a Small Language Model to Play Codenames

held on October 1, 2026, at Devon Energy Hall, The University of Oklahoma, and completed every stage of the build:

✓Implemented the rules of Codenames in Python
✓Generated a verified training dataset with an embedding oracle and a verifier, with no human labels
✓Fine-tuned Google’s Gemma 4 with QLoRA (Unsloth) on an INQUIRE Lab GPU
✓Evaluated the fine-tuned model on held-out game boards
✓Deployed an AI agent of their own, “ruijing zhang”, in the Codenames arena

Arash AhmadiArash AhmadiPresident, LLM Engineering ClubINQUIRE Lab, The University of Oklahoma

Certificate ID STT3XNYWorkshop date October 1, 2026Issued October 2, 2026Verify at inquirelab.ai/c/STT3XNY

About this workshop

In a two-hour session, students built a Codenames-playing language model from scratch: they wrote the game's rules, turned words into embeddings, generated their own training data with a clue oracle and a verifier, fine-tuned Google's Gemma 4 with QLoRA on INQUIRE Lab GPUs right in the browser, measured the result on boards the model had never seen, and then played against it.