Hani Mir
Previously leading model finetuning at Character.ai
Scaling LLMs to 20k+ QPS, reducing inference costs by 75%
10+
Years Experience
70B
Parameter Models Scaled
75%
Inference Cost Reduction
20k+
QPS at Scale
LLM Infrastructure
Scaling 70B parameter models to 20k+ QPS, implementing parameter-efficient fine-tuning, and building real-time inference pipelines.
Distributed Systems
Building resilient infrastructure managing O(100k) securities, achieving 4 nines reliability, and scaling from startup to enterprise.
ML Optimization
Driving 3x GPU utilization improvements, 75% cost reductions, and implementing safety alignment with measurable impact.