Devashish Lal
Machine Learning Systems Engineer at Meta
Los Angeles, California, United States
Summary
👤
Senior🎓
Top SchoolDevashish Lal is a Machine Learning Systems Engineer with 8 years of experience specializing in high-performance LLM inference, compiler fusion, and hardware–software co-design, currently building production-grade inference at Meta. He has a strong track record of upstreaming open-source performance improvements—reducing kernel counts, massively improving memory efficiency, and stabilizing CI that runs 80+ LLMs nightly—and co-leads the sgl-fusion stack powered by torch.compile, dynamo, and inductor. His work spans kernel fusion, graph-level transforms, custom accelerator backends, and MLPerf-style benchmarking, turning novel model architectures (MLA, linear attention, MoE) into memory-efficient, deployable systems. Past projects include SGLang and Flashinfer on the Rivos stack, 3D and NeRF research for XR at Samsung Research America, and large-scale memory reductions in Unreal Engine demos at USC ICT. He is an active open-source collaborator (including CI/DevOps improvements on the popular game-ci/unity-builder) who favors reusable infrastructure and practical performance wins that close the gap between model research and real-world serving.
8 years of coding experience
6 years of employment as a software developer
Masters Computer Science, Masters Computer Science at University of Southern CaliforniaBachelor of Technology - BTech Computer Science, Bachelor of Technology - BTech Computer Science at Manipal University Jaipur
Stackoverflow
101reputation
72kreached
1answer
1question
Github Skills (16)
unity3d10
build210
cil10
github-ci10
githubaction-workflow10
dockers9
bash9
docker9
gradle6
react6
nrwl6
javascript6
electron6
intellij-idea6
ecmascript6
Programming languages (11)
Github contributions (5)
game-ci/unity-builder
Oct 2020 - Oct 2020
Build Unity projects for different platforms
Role in this project:
DevOps Engineer
DevOps EngineerContributions:2 reviews, 9 commits, 2 PRs in 4 days
Contributions summary:Devashish's contributions primarily focused on modifying the build process and CI/CD pipeline for the Unity project. They refactored the build scripts, optimized the use of the unity-editor alias, and removed redundant calls. Additionally, the user updated image tags and added the linux platform to the build process. These changes indicate a focus on improving the build environment and ensuring compatibility across different platforms.
game-ciplatformsunity-actionsunity-builderunity
BLaZeKiLL/UE5VoxelTutorial
Jun 2021 - Aug 2022
A collection of voxel mesh generation algorithms
Contributions:47 commits, 5 PRs, 36 pushes in 1 year 2 months
meshminecraftunreal-enginemesh-generationvoxels
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.