---
title: "Technical Report: The Fleet Strategy — Distributed Inference Across Heterogeneous Consumer Hardware — Hayula Research"
authors:
- Hayula AI Lab
papers:
- paper.md
tags:
- fleet
- distributed-inference
- heterogeneous-hardware
license: mit
---

# Technical Report: The Fleet Strategy — Distributed Inference Across Heterogeneous Consumer Hardware — Hayula Research

**Hayula AI Lab**

## Abstract

This technical report describes Hayula Labs' fleet strategy for orchestrating AI inference across a heterogeneous collection of consumer-grade machines. We detail the architecture, operational metrics, and cost analysis of a production deployment spanning five machines in Kuwait (Mac Studio M2 Ultra 192GB, Acer Nitro V14 with RTX 4050 6GB, Framework Desktop, legacy workstations, NAS) with a unified routing layer. Over 90 days of operation, the fleet processed 3.9M inference requests with 99.4% u

## Files

| File | Description |
|------|-------------|
| `paper.md` | Full paper (Markdown) |
| `README.md` | This model card |

## Citation

```bibtex
@techreport{hayulalab2026fleetstrategy,
    title={Technical Report: The Fleet Strategy — Distributed Inference Across Heterogeneous Consumer Hardware — Hayula Research},
    author={Hayula AI Lab},
    year={2026},
    url={https://huggingface.co/hayulalab/fleet-strategy-paper}
}
```

---

*hayulalab — Open Source AI Research*