Skip to content
View ryoari's full-sized avatar
4 followers · 6 following

Achievements

Achievement: Pull Sharkx2Achievement: Pair ExtraordinaireAchievement: YOLO

Achievements

Achievement: Pull Sharkx2Achievement: Pair ExtraordinaireAchievement: YOLO

Highlights

  • Pro

Block or report ryoari

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
ryoari/README.md

Rajat Roy (@ryoari). AI systems. Evidence included. A cyan computational sculpture on a graphite background.

OpenVerifiableLLM maintainer · GSoC 2026 @ AOSSIE
Master of Computer Science student · University of Sydney

Website Source code Experiment evidence

I build tools for verifiable training and autonomous research.

I care about the evidence behind a result: where the data came from, how the system ran, and whether someone else can reproduce it.

Featured work · OpenVerifiableLLM

Language-model training that people can inspect and check. My work focuses on deterministic execution, data provenance, checkpoint verification, and reproducible releases.

23.1M model parameters. All 37 training segments replayed by the project on a second RTX 5090.

The project reproduced all 37 training segments bit for bit on a second RTX 5090. These checks cover the training computation on matching GPU hardware.

Source code ↗  ·  Method and results ↗  ·  My contributions ↗

Research interests

  • Autonomous research. Agents that choose useful experiments, run them, and retain the evidence.
  • Reliable execution. Observable runs, controlled actions, and recovery from failure.
  • Complex systems. Emergence, information flow, and artificial life.

Working tools

Python PyTorch safetensors uv GitHub Actions

Small experiments. Clear failure records. Results others can reproduce.

Outside code, I write speculative fiction and read fantasy manhwa.

Pinned Loading

  1. Verifiable-LLM-Baseline Verifiable-LLM-Baseline Public

    Supporting work for my proposal in GSOC'26 for https://github.com/AOSSIE-Org/Info/blob/main/GSoC-Ideas/2026/LLM.md

    Python

  2. springboardmentor553-maker/Autonomous-Cognitive-Engine-for-Deep-Research-and-Long-Horizon springboardmentor553-maker/Autonomous-Cognitive-Engine-for-Deep-Research-and-Long-Horizon Public

    Python 1 4

  3. modis-landsat-upscaling modis-landsat-upscaling Public archive

    A project that aims to upscale daily modis surface reflectance (500m) into 8 day landsat surface reflectance (30 m) via Ai

    Python 1

  4. Code.Similarity.Index Code.Similarity.Index Public archive

    Plagarism Detection

    PowerShell 1

  5. tower_of_chance tower_of_chance Public archive

    Python 1

  6. Noesis Noesis Public

    an hierarchial tree of current ai research