Joe McKenna

AI Researcher

I work on post-training for code generation and efficient inference.

Samples

ModelParts: Model-Grounded Tasks and Roofline Rewards for Training Kernel Agents on Emerging Accelerators
ICLR 2027 (submitted)
CommunityKV: Efficient Long-Context Decoding via Graph Partitioning
NeurIPS 2026
Semantic Complexity in End-to-End Spoken Language Understanding
Interspeech 2020

Experience

Sr. Applied Scientist — Amazon AGI
Nova Foundation Models
Sr. Software Engineer — Google Cloud
Enterprise Search and Agents
Postdoctoral Fellow — NIH
Laboratory of Biological Modeling
PhD, Mathematics — Florida State University