Skip to content
View tornikeo's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report tornikeo

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.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

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

Report abuse
tornikeo/README.md

About

I'm Tornike Onoprishvili (aka tornikeo). I specialize in AI, GPUs, and the Cloud. If you have a cool research idea or project please reach out by email at tornikeonoprishvili [AT] gmail [dot] com. I'm always happy to help out and collaborate.

Projects

  • I built and maintain the fastest exact Cosine Similarity calculation tool. It's a simple, unoptimized CUDA kernel that calculates mass spectrum Cosine Similarity and Modified Cosine Similarity scores. The 1700x speedup comes from the typically massive memory bandwidth of GPUs. No shared memory or highly optimized memory access is used. I hope to optimize this further at some point.
  • I've designed key parts of an upcoming MS/MS Foundation model, SpectruMS. The paper is a work in progress.
  • I built a finite-difference time-domain simulation tool for fully all-optical logic gates. These "optical gates" are made of a special glass that changes its refractive index $n$ as the power of the laser increases. This allows us to make a fully optical AND gate. The simulations also show how the all-powerful NAND gate can be built. All simulation code is in MATLAB.

Pinned Loading

  1. PangeAI/SimMS PangeAI/SimMS Public

    GPU-Accelerated Cosine Similarity for Tandem Mass Spectrometry

    Jupyter Notebook 13 3

  2. optical_nand optical_nand Public

    FDTD simulation of optical NAND inside a nonlinear photonic crystal

    MATLAB 2 1

  3. tblite-gpu tblite-gpu Public

    GPU accelerated kernel for extended tight-binding

    Cuda 2