Molly OS - Whitepaper / Preprint
Daniele Trovato - Core Labs R&D - v1.0 - June 2026
Molly OS: A Model-Agnostic Inference Orchestration Layer for On-Device and Federated Inference.
A sovereign orchestration layer that routes each request across heterogeneous execution
targets (on-device, LAN, cloud, external API), serves many domain-specialist LoRA adapters
over a shared quantized base, enforces on-device-first data sovereignty, and continuously
specializes via distillation. Evaluation across ~100 domains (neutral LLM judge, 115-panel
probe) shows orchestration improves output quality over the unspecialized base.
Links
Read the paper
- On this page (renders with figures):
molly_os_whitepaper_en.md - molly_os_whitepaper_es.md - molly_os_whitepaper_it.md
- PDF (offline / print, figures embedded):
molly-os-whitepaper-en.pdf
- Branded rendered version (live): EN - ES - IT
Architecture figures are embedded as SVG in the .md views (fig_1.svg ... fig_5.svg)
and rasterized into the PDF. The 69 references are verified against the arXiv API, with
official venue links for the two non-arXiv works.
License
Status: PUBLIC preprint. This repository contains three artifact classes, each
licensed separately:
| Artifact | License | What it allows |
|---|
| Whitepaper, translations & figures | CC-BY-4.0 | Share, quote, translate and build upon with attribution (incl. commercial) |
| Source code | Apache-2.0 | Use, modify and distribute, including commercially |
| Model adapters / weights | CC-BY-NC-4.0 | Share and adapt with attribution; no commercial use |
You may read, cite, quote, translate, and reuse the whitepaper with attribution. See
LICENSE. The orchestration
service is the commercial product; for commercial use of
the adapter weights, contact
info@corelabsgroup.com.
How to cite
1@techreport{trovato2026mollyos,
2 title = {Molly OS: A Model-Agnostic Inference Orchestration Layer for On-Device and Federated Inference},
3 author = {Trovato, Daniele},
4 institution = {Core Labs R&D},
5 year = {2026},
6 month = {6},
7 note = {Preprint, v1.0}
8}
Contact (corresponding author): Daniele Trovato - Core Labs R&D -
info@corelabsgroup.com
(c) 2026 Daniele Trovato / Core Labs R&D. Contact:
info@corelabsgroup.com
