Skip to content
View kuntal-devrat's full-sized avatar
🎯
Focusing
🎯
Focusing
21 followers · 29 following

Achievements

Achievement: Pull Sharkx3Achievement: Pair Extraordinairex3Achievement: YOLOAchievement: Quickdraw

Achievements

Achievement: Pull Sharkx3Achievement: Pair Extraordinairex3Achievement: YOLOAchievement: Quickdraw

Block or report kuntal-devrat

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. 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
kuntal-devrat/README.md

Devrat Kuntal

AI Architect & High-Performance Systems Engineer
Architecting GPU compute pipelines, zero-cost abstractions, and deterministic AI agent runtimes.

Profile Views Support on Ko-fi PyPI npm Repositories

⚡ Focus

  • Accelerated Computing: High-throughput GPU pipelines using WebGPU (WGSL), zero-cost Rust abstractions, and vectorized SIMD kernels (AVX-512, AVX2, NEON).
  • Autonomous Runtimes: Deterministic execution layers and zero-latency tool-dispatch runtimes for local and in-browser AI agents.
  • Compilers & Systems: Declarative query transpilers, sub-millisecond polyglot runtimes, and hardware-directed IEEE 754 precision libraries.

🚀 Featured Projects

Project Description Stack Link
fuzzgpu GPU-accelerated fuzzy string matching & alignment (6.3× faster than RapidFuzz, zero CUDA deps) Rust WebGPU SIMD pip install fuzzgpu
torchburn Hardware-agnostic, universal GPU-first PyTorch compilation backend in Rust Rust PyTorch WGSL Source →
PipeQL Pipelined, injection-safe query compiler (transpiles to PostgreSQL, DuckDB, SQLite) Rust Compilers Source →
alloy Hyper-optimized systems runtime (<1.5ms cold starts, zero-copy interop, actor concurrency) Rust C Python Source →
py-nerve Deterministic desktop automation & agent tool dispatch layer with sensory feedback Python AI Runtimes Source →
precise-numpy Interval arithmetic arrays with hardware-directed IEEE 754 rounding and SIMD kernels Python PyO3 AVX-512 pip install precise-numpy
rezon In-browser autonomous AI agent runtime for static HTML without server backends TypeScript Web AI npm install rezon
qyro High-concurrency polyglot code execution engine inside isolated micro-environments TypeScript Docker Source →

🛠️ Core Stack

  • Languages: Rust · Python · C++ · TypeScript · WGSL / GLSL
  • Compute & Systems: WebGPU (wgpu) · SIMD (AVX-512 / NEON) · PyO3 · Rayon · WebAssembly
  • AI & Infrastructure: PyTorch · NumPy · FastAPI · Docker · PostgreSQL · DuckDB

☕ Support My Work

If you find my open-source tools, compilers, or GPU compute libraries helpful, consider supporting my work on Ko-fi. Contributions support hardware and cloud compute resources for testing and benchmarking.

Support on Ko-fi

📊 Activity & Analytics

GitHub Stats Top Languages
Devrat Kuntal • @kuntal-devrat • Support on Ko-fi

Pinned Loading

  1. mango mango Public

    An ultra-lightweight, memory-efficient web browser and rendering engine written from scratch in pure Rust.

    Rust 7 1

  2. PipeQL PipeQL Public

    Pipelined, injection-safe polyglot query language. Compiles to SQL for Postgres, SQLite, DuckDB, MySQL.

    Rust 1

  3. py-nerve py-nerve Public

    High-speed, deterministic desktop automation framework & AI agent execution layer for Python (pip install dexflow)

    Python 1

  4. fuzzgpu fuzzgpu Public

    Hardware-accelerated fuzzy string matching & sequence alignment — cross-platform, no CUDA required

    Rust

  5. torchburn torchburn Public

    Hardware-Agnostic, Universal GPU-First PyTorch Compilation Backend in Rust (Zero CUDA)

    Rust

  6. lgpu lgpu Public

    Ultra-lightweight, bindless-first cross-platform GPU abstraction

    Rust 7 1