Avi Fenesh

With over three years at Amazon Web Services (AWS), I design and build high-performance software systems for large-scale, in-memory data stores, which are used by millions. My work centers on optimizing latency, throughput, and resource efficiency through Rust, Java, Node.js, and many other languages — with deep use of performance critical methodologies and OS internals, such as zero-copy, FFI, shared memory, runtime internals, async and multi threading and other related mechanisms. I’ve led initiatives around module persistence, logical correctness, crash recovery, and cluster replication, shaping resilient distributed systems that power critical workloads.Beyond my primary engineering role, I maintain and lead the Valkey Glide open-source client library, driving community engagement, contributors relations and onboarding, and ecosystem integration. I believe in practical engineering, transparency, and shared growth — helping teams and developers build better systems while learning from one another. My passion lies in advancing core infrastructure, pure curiosity, and the beauty rooted in logical and novel solutions — from kernel-adjacent projects and storage engines to distributed caching and database internals — while keeping software approachable, performant, and dependable.Outside work I run applied ML research on my own hardware: training compact speculative-decoding draft heads, MoE pruning and healing, and NVFP4 quantization — measured end to end in bw24, a from-scratch Rust + CUDA inference engine I built for exactly that. Working papers and evidence ledgers: avifenesh.ai