Kristin Ashcraft

Gmail Google Apps Microsoft Office 365 Mobile Friendly Greenhouse.io Remote AI Render Circle Sigma Airtable Lightspeed Basis

computer hardware manufacturing high throughput llms ai models hardware design performance optimization cost efficiency training inference computing power performance-per-dollar low latency large models dense models mixture of experts scale-out performance training flops interconnect technology transformer models 10t-class models gpt-4 training memory bandwidth kv cache distributed systems simultaneous users multihost partitioning checkpointing deterministic data loading parallel computation explicit memory usage model training research codebase ml frameworks code simplicity resource efficiency sharding profiling debugging ml inference optimization hardware accelerators chip architecture data loading fsdp tensor parallelism inference costs memory footprint fine-tuning model scalability open source computer software information technology & services