Compliance assessments across 15+ frameworks, grounded in the customer's own policies.
Hassan works on the platform's backend and AI workflows as a contract engineer, through Xeltec, since 2025.
- Client
Arion Networks- Sector
- Compliance AI
- Scope
- Python · Multi-LLM RAG · Supabase
- Live at
- iso.arionetworks.com

ProofArionComply
ArionComply replaces spreadsheets, consultants, and disconnected tools with one platform for US and EU compliance frameworks, built for small businesses, MSPs, and the under-resourced organisations that incumbents overlook.
That a gap assessment can be a conversation that remembers what it was told, grounded in the documents a company already has.
Compliance answers have to be traceable, so every step keeps its record: prior answers loaded before the model speaks, validated facts extracted after, and the AI workflows themselves versioned and observed.
- 01
Retrieval over policies
retrieve(policies, question)Multi-LLM retrieval over customer policy documents, with embeddings and vector search, so answers come from the company's own words.
- 02
Stateful assessments
assess(answer) → factsEach conversation loads the prior answers before inference, extracts validated facts after every response, and saves its progress.
- 03
Self-hosted inference
llama.cpp · serve()Inference on self-hosted models, with an AIOps dashboard for versioning and observing the AI workflows.
- 04
The compliance core
log · register · notifyContributions across event logging, the incident register, breach notification, and MFA, on Supabase with Postgres row-level security.
Figures
fig. 2, a sketch
Approaching its demo and MVP.
Next exhibit · FWaste management
IntellaWaste


