Pinned Loading

1. horizon-agent horizon-agent Public

   Evidence-first multi-agent demo for renewable-asset investigation on synthetic data — LangGraph routing, cited RAG, inspectable execution traces, local LLM inference.

   Python

2. grpo-training-lab grpo-training-lab Public

   Single-GPU GRPO trainer for LLMs: LoRA fine-tuning, Hugging Face & vLLM rollouts, resumable checkpoints, and a memory-efficient Triton selected-logprob kernel. Tested on Qwen2.5 (0.5B/1.5B) on an 8…

   Python

3. evals-judge-calibration evals-judge-calibration Public

   Auditable, fully local agent evaluation: stateful benchmark, executable graders, LLM-as-judge calibration, blinded annotation tooling, repeated-run statistics, and failure analysis.

   Python

4. Spaceship-Titanic Spaceship-Titanic Public

   Data analysis and modelling for the Kaggle Spaceship Titanic competition.

   Jupyter Notebook 1

5. Student-Behavior-Analysis-and-Course-Recommendations Student-Behavior-Analysis-and-Course-Recommendations Public

   Analyzes student behavior and engagement in online learning, with a rule based and machine learning enhanced course recommendation system.

   Jupyter Notebook 1 1

6. voxel-mcp voxel-mcp Public

   MagicaVoxel MCP Server - 107 tools for procedural voxel modeling

   TypeScript 3