Building a Miniature AI Factory: Post-Training, Inference Engines, and Provider Systems

Tanvir Ahmed

doi:10.5281/zenodo.22886076

RResearcharAcademic radar for researchers

Feed Venues Groups Network Inbox Profile

Your academic radar

One personalized feed of papers, calls for papers, collaborators and datasets — plus research groups and paid-in-points study recruitment.

Create account Log in

Top venues

Browse all

Loading…

Building a Miniature AI Factory: Post-Training, Inference Engines, and Provider Systems · Researchar

Back to feed

Paperopenalex

ZCE

Building a Miniature AI Factory: Post-Training, Inference Engines, and Provider Systems

Tanvir Ahmed · Zenodo (CERN European Organization for Nuclear Research) 2026 · 2026

DOI: 10.5281/zenodo.22886076

0citations

Explainable Artificial Intelligence (XAI)Multimodal Machine Learning ApplicationsMachine Learning and Data ClassificationComputer scienceInference

Citations by source

Synced 2026-09-24T02:30:51.189939Z

0OpenAlex · primary0Semantic Scholar0Crossref

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Abstract

This paper is the standalone technical-report edition of the Mini AI Stack study. It explains a compact, reproducible chain spanning supervised fine-tuning, REINFORCE, PPO, RLOO, GRPO, workload-dependent inference engines, and an OpenAI-compatible provider boundary. A three-seed, 21-cell template-disjoint campaign found 4/128 held-out exact matches for every tested adapter. Canonical links: project article; source repository; LoRA adapter; software release.

Building a Miniature AI Factory: Post-Training, Inference Engines, and Provider Systems

Citations by source

Abstract

Discussion (0)