Hi, I'm Christina Norman

AI/ML contributor focused on LLM inference optimization and tooling. Active contributor to vLLM and Hugging Face Transformers.

Background: 20+ years in games — Lead Combat Designer on Mass Effect 1-3 at BioWare, Design Lead at Riot Games (League of Legends, Wild Rift), Founder of Elodie Games.

Current Focus

Building and improving LLM infrastructure—particularly GGUF model support, memory optimization, and inference performance for next-gen hardware like NVIDIA Blackwell.

Also exploring AI-assisted game development workflows, combining game design expertise with modern AI tooling.

Open Source Contributions

vLLM (High-throughput LLM Inference Engine)

Merged:

• fix(gguf): Disable bfloat16 for GGUF on Blackwell
• fix(shm): Add memory barriers for cross-process shared memory
• Skip generation config fallback for GGUF to prevent hang

In Progress:

• GGUF dtype auto-selection for Blackwell GPUs
• Gemma2/Gemma3 GGUF weight loading fixes
• Nemotron-H rotary positional embeddings
• Lazy tokenizer init to prevent semaphore leaks

Hugging Face Transformers

• Add attn_logit_softcapping to Gemma2/Gemma3 GGUF config

Claude Code

Active contributor to Anthropic's Claude Code CLI—filing feature requests and bug reports to improve the AI coding agent experience.

Tech Stack

GitHub Stats

Connect

• LinkedIn: therealchristina
• X/Twitter: @truffle
• Location: Austin, Texas

---

_{University of Waterloo · Mass Effect · League of Legends · Wild Rift · AI/ML · LLM Inference}