Vertical-Driven Architectures
Collection
Physical AI system designs by CodeNinja, reasoned on Praxis, made living with Hyper Ontology. • 89 items • Updated
Datasets:
Modalities:
Text
Formats:
csv
Languages:
English
Size:
< 1K
Tags:
physical-ai
sovereign-ai
united-states
supply-chain-and-logistics
self-improving-ai-system
ontologies
License:
choice stringlengths 7 30 | picked stringlengths 4 185 | why stringlengths 27 284 |
|---|---|---|
Reasoning model | GLM 5.3, filed at 753B parameters, FP8 (753 GB of weights), GLM-5.3 bespoke license, served on vLLM on one node in the operator's own facility in the United States | The model the requirement names; its license permits commercial use, fine tuning and redistribution, so the operator holds its own copy with no per token obligation; carries the triage and recovery drafting, the command center's cited conversational interface and ontology maintenance |
Site tier parsing model | Qwen3.8 27B, dense, Apache-2.0, served at FP8 (about 31 GB of weights) on one 48 GB PCIe card in an air cooled server | High volume thread and note parsing over years of archived email and chat and the live threads, a throughput workload kept off the frontier node so coordinator sessions keep their concurrency; about 13 GB of KV ceiling on the card |
Embedding model | Qwen3-Embedding 0.6B, Apache-2.0, about 1.2 GB at BF16 and about 0.6 GB at 8 bit, sharing the site tier server, 32K context window | Embeddings over threads, notes and resolutions so retrieval finds the past exception that matches this week's; permission aware, returning only what the asking coordinator could already open |
Frontier hardware | 141 GB HBM GPU class, eight GPUs per node, one node (1,128 GB) | Holds 904 GB (753 GB of FP8 weights times the 1.2 planning factor) with about 280 GB of KV cache ceiling at the 0.92 utilization the sizing rule allows; about 15 kW of rack power per node with margin, 10 RU, 10 to 35 C inlet |
Frontier footprint sizing rule | Frontier Model GPU Footprint Sizing Rule (vLLM, SGLang): parameters times one byte at FP8 times 1.2 | Turns filed parameter counts into a memory requirement before any workload is placed |
Node count rule | GLM 5.3 node count by GPU class | One node of eight 141 GB GPUs; two nodes of the 80 GB class or BF16 serving (about 1.5 TB) rejected because they add a second node joined over an interconnect |
Site tier hardware | 48 GB PCIe card class, one card, in an air cooled PCIe server the facility's server room already takes | Fits the parsing model's FP8 weights with KV ceiling to spare and carries the embedding model beside it without a second server |
Serving runtime | vLLM | Serves the FP8 weights within the KV cache ceiling the sizing reserves |
Registry | MLflow Model Registry on site | Pinned model images; serving restarts from pinned images with a rollback exercised once under observation; the desk never runs on an unobserved weight change |
Event backbone | Apache Kafka (KRaft mode) | Ordering by offsets and per stream sequence numbers with each source's clock offset recorded, at least once delivery with idempotent projection, buffering in the retained log, replication inside the facility |
Time series store | InfluxDB 3 Core | Tracking telemetry under the object model |
Orchestration | Kubernetes | Containers for the services and the serving tier inside the facility |
Identity | Keycloak | Federated identity for coordinators and permission aware retrieval scoped by each shipper's data terms |
Observability | Prometheus | Observes the running system |
Sensing | None; the existing carrier tracking and visibility feeds are read | Trucks are seen through feeds the operator already owns; no camera, positioning hardware or edge box is built |
Pattern: memory | Memory loops over confirmed outcomes, nothing trained, fine tuned or distilled into weights | The requirement's own words: better every week without training a model on the fly; memory is auditable where weights are not |
Pattern: approval | Every draft confirmed by a named coordinator; one gated write path into the scheduling tool | Confirmations logged as ground truth, with known answer probes and weekly sampling against rubber stamping |
Pattern: baseline | The pilot lane group's exception measures are recorded before any coordinator sees a draft | Improvement is measured against evidence, not impression; the phase 1 gate reads time to resolve and repeat exceptions against that baseline |
Pattern: boundary | The operator's own hardware in its own facility in the United States; no inference call, retrieval or draft leaves the building; only approved carrier messages and shipper notices cross | Shipper and carrier data never leave the operator's hardware, open weights only, no third party model service |
A Self-Improving AI System for Carrier Exception Handling in the United States: ontology and model register
The object model and the model and equipment register from A Self-Improving AI System for Carrier Exception Handling in the United States, an open reference architecture by CodeNinja Atoms. Self-improving AI system for carrier exception handling in the United States: one freight ontology, three drafting agents, GLM 5.3 on the operator's own node. Part of the Vertical-Driven Architectures series; every design in the series is also a row in the cumulative dataset https://huggingface.co/datasets/CodeNinjatools/vertical-driven-architectures.
- Read the paper: https://codeatoms.ai/carrier-exceptions-us/
- DOI: https://doi.org/10.5281/zenodo.24048585
- Source files and PDF: https://github.com/muhammadumar89/codeninja-research
- Live view: https://huggingface.co/spaces/CodeNinjatools/carrier-exceptions-us
Files
| File | What it holds |
|---|---|
objects.json |
16 typed objects (Shipper, Order, Shipment, Stop, Appointment, Carrier, Driver, Tracking event, Exception, Exception type, Action, Resolution, Standing fix, Claim, Scorecard entry, Exception coordinator), with 13 typed links. Format hyper-ontology/1: designed with CodeNinja Praxis, implemented with Hyper Ontology. |
models.csv |
The model and equipment register: every model and hardware class with its license and the reason for each choice (19 rows). |
How to use it
Load objects.json with the Hyper Ontology loader (pip install hyper-ontology, see https://codeatoms.ai/developer/) or read it as plain JSON. The operator is described by class, never by name.
Made with CodeNinja Praxis and Hyper Ontology. One of the CodeNinja Atoms designs for self-improving AI system. CC BY 4.0.
- Downloads last month
- 35