Skip to content
View cheese-cakee's full-sized avatar
♨️
high cortisol engineering
♨️
high cortisol engineering
163 followers · 34 following
  • 03:51 (UTC +05:30)

Achievements

Achievement: Pull Sharkx3Achievement: YOLOAchievement: StarstruckAchievement: Pair ExtraordinaireAchievement: Quickdraw

Achievements

Achievement: Pull Sharkx3Achievement: YOLOAchievement: StarstruckAchievement: Pair ExtraordinaireAchievement: Quickdraw

Highlights

  • Pro

Organizations

@hiero-ledger

Block or report cheese-cakee

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
cheese-cakee/README.md 

╭─ cheesecake@github ─────────────────────────────────────────────────────────────╮
│                                                                                 │
│  profile   CS undergrad | systems, inference, dev tools                         │
│  trust     LFX Mentee @hiero-ledger, ex contract inference engineer @lucebox    │
│  impact    merged PRs in vLLM, Ceph, CRIU, Rizin, Prometheus, lucebox           │
│  strength  C/C++ systems, GPU inference (CUDA, HIP), model serving              │
│  workflow  exact-output checks, matched benchmarks, clean commits               │
│                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────╯

my work so far

  • Lucebox inference engine · contract inference engineer, 2026
    24 merged PRs in a C++ CUDA/HIP inference engine: DeepSeek-V4 MoE prefill across two AMD GPUs (2.02x on a 401-token prompt, byte-identical output), a fused MoE combine kernel, HIP wave64 correctness, and a CUDA-graph replay fix.

  • LFX Mentorship 2026 · Hiero · ongoing
    Building sdk-automations, one GitHub App that runs contributor-facing automation for the Hiero SDK repositories from repository-owned config: signature-verified webhooks, crash-safe SQLite processing, and checks that stop it acting on stale state. Before the mentorship: CI bots and tests in the Python and C++ SDKs.

  • Systems and infrastructure · merged fixes in Ceph (RGW), CRIU, Rizin, and Prometheus.

some stuff I know

C C++ Go Python Bash Git CUDA HIP/ROCm TypeScript CMake Docker Linux

github stats

Pinned Loading

  1. Benchmarking-GEMM Benchmarking-GEMM Public

    Step-by-step C++17 GEMM optimization: ten kernels from naive ijk to AVX2 and OpenMP. Peaks at 490 GFLOPS at 2048x2048 on i5-13450HX.

    C++ 19

  2. overtrust overtrust Public

    overtrust is a CLI binary written in C++. no LLMs, no AI, no cloud. it uses deterministic rules, regex pattern matching, shannon entropy analysis, and manifest parsing to find what has access to yo…

    C++ 8

  3. hiero-hackers/sdk-automations hiero-hackers/sdk-automations Public

    A on/off, service-neutral Github App for contributor-facing automations. Safety first by construction, flexible config, contributor-friendly

    TypeScript 4 10

  4. bonsai-inference-lab bonsai-inference-lab Public

    Bonsai 2 27B PTQ1 inference on a 6 GB RTX 4050: AVX2 patch, benchmarks, and reproducible WSL2 case study

    PowerShell 1

  5. Unix_networking Unix_networking Public

    A collection of POSIX networking programs

    C 5