Home Original page

monigarr - Overview

MoniGarr

Foundational AI Systems • Language Technology • AI Governance

MoniGarr is an independent AI research and engineering practice focused on the design, stewardship, and long-term governance of AI systems operating in high-risk language, cultural, and institutional contexts.

This work predates modern large language models and spans pre-LLM conversational systems, custom NLP pipelines, reinforcement learning systems, and contemporary model architectures.

The emphasis is durability over scale, judgment over output, and institutional trust over visibility.

Research Focus

Indigenous & Low-Resource Language AI

  • Custom tokenizers and language models for polysynthetic languages (Kanien’kéha)
  • Morphology-aware NLP pipelines
  • Data sovereignty and language stewardship–aligned system design

Reinforcement Learning Systems

  • Modernized SAC agents for Gymnasium-compatible environments
  • Reproducible RL research pipelines with evaluation, logging, and model cards
  • Applied RL for environments where failure modes matter

AI Infrastructure & Tooling

  • Research-grade ML pipelines
  • Synthetic data workflows
  • AI-first developer systems for edge-case domains

Technical Stack

Python • PyTorch • TensorFlow • Hugging Face • Gymnasium • RLlib
MLflow • Jupyter • Docker • GitHub Actions • Unity • Blender

Tooling choices prioritize reproducibility, interpretability, and long-horizon maintenance.

Selected Work & References

Engagement Model

Work is conducted through:

  • Sponsored research
  • Fellowships
  • Selective advisory or stewardship roles

This repository reflects research artifacts and system designs rather than productized outputs.

Contact

https://www.monigarr.com
https://www.linkedin.com/in/monigarr

Inquiries are evaluated for long-term alignment and institutional fit.