Skip to content
View bgub's full-sized avatar
💻
Improving the web
💻
Improving the web

Sponsors

@generaltranslation

Block or report bgub

Report abuse

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

Report abuse
bgub/README.md

Ben Gubler

Looking for a project that isn't here? Check out @nebrelbug, my old account where I store archived/inactive projects.

  • Student at BYU w/ a dual major in ACME (Applied and Computational Mathematics) and Arabic
  • Currently interning at Vercel
  • Previously conducted research on AI alignment + mech. interp. at BYU's Perception, Control, and Cognition Lab

Projects - AI/ML/GPU

  • tokka - Toolkit for training BPE tokenizers on custom data splits (tokenizers, datasets, mosaicml-streaming)
  • tokka-bench - Benchmark and compare tokenizers across many languages (tokenizers, streamlit)
  • hf_to_mds - Convert HuggingFace datasets to MosaicML Streaming format (MDS) for efficient cloud-based training (datasets, mosaicml-streaming)
  • gom - CLI tool to monitor GPU usage across Docker processes (pynvml)
  • mokka (WIP) - Toolkit for creating SLMs easily, targeted towards researchers (PyTorch, OmegaConf, WandB)
  • shade (WIP) - PyTorch-like computation library for TS/JS using WebGPU (Deno, WGSL)

Projects - Other

  • Eta - Fast embedded JS template engine. Widely used with 1.5K+ GitHub stars and ~1M weekly downloads. (TypeScript, microbundle, Jest, GitHub Actions)
  • TinyLingo - Language learning app for improving comprehension. Listen to authentic native podcasts with AI-generated transcriptions and translations. (Next.js 15, Convex, Clerk, shadcn/ui, OpenRouter)
  • bengubler.com - My personal website and blog where I write about AI, distributed training, and language learning. (Next.js 15, Tailwind, MDX)
  • nix-macos-starter - Starter Nix config for macOS (nix-darwin, home-manager, homebrew, mise)
  • Decline App - PWA for practicing Czech, Slovak, and Russian declension patterns. (Next.js, Tailwind, PWA)
  • npm-to-yarn - Converts between npm/Yarn/pnpm/bun CLI commands. (TypeScript, Jest)
  • Open source contributions - Various TypeScript and web development projects on GitHub.

You can find more of my work at bengubler.com, where I post projects and write about AI, distributed training, and language learning.

Interests

  • Languages - Fluent in English and Czech; conversational in Slovak; learning Russian and Arabic; have also studied Esperanto, toki pona, and Spanish
  • Machine Learning - Large language models, AI alignment, mechanistic interpretability, distributed training
  • Web Development - TypeScript, React ecosystem, performance optimization, developer tooling
  • Outdoor Activities - Spelunking, rock climbing, backpacking, pickleball
  • Literature - Classic literature, poetry, fantasy, sci-fi

nebrelbug

Pinned Loading

  1. eta eta Public

    Embedded JS template engine for Node, Deno, and the browser. Lighweight, fast, and pluggable. Written in TypeScript

    TypeScript 1.6k 81

  2. tokka-bench tokka-bench Public

    benchmarks for LLM tokenizers

    Python 14 1

  3. npm-to-yarn npm-to-yarn Public

    Convert npm CLI commands to yarn, and vice versa

    TypeScript 37 13

  4. bengubler.com bengubler.com Public

    My personal website, built with Next.js 15

    MDX 5 6

  5. ts-base ts-base Public

    Starter TS library template. Vitest, Biome, tsdown, CI publishing, JSR, Deno, etc.

    TypeScript 8

  6. nebrelbug/llm_trainer nebrelbug/llm_trainer Public

    Comprehensible scripts to instruction-tune a LLaMA model

    Python 2