Data and compute under Greek and European control.
Five layers, each one measured. Products are built on top, so an improvement in the core improves every call, every booking and every answer.
Turns a request or a “hire” into an App Manifest (entities, workflows, skills, channels, policies and connectors) and provisions it. Publishing is provisioning.
Voice, WhatsApp, web apps and telephony (SIP) all reach the same Brain, so a booking made by phone is visible in chat.
An employee is a role plus skills plus policies. Each skill is a typed contract with deterministic code, tests and applicability conditions, signed and evaluated with and without the skill.
Your business as typed entities, relations and rules in a bitemporal store. A constraint gate checks every proposed action, and every decision is written to an append-only, hash-linked ledger. The model proposes; the rules decide.
Open-weight models on vLLM with FP8 and prefix caching, on GPUs in Greek data centres. Confidence-based routing between models; every failed trace becomes a skill fix or a fine-tuning example.
The rule applies to every product served from the Hellas Inference core. AI Employee runs today on EU cloud with third-party model APIs and moves onto the Greek core in 2027; the rule applies to it from that migration.
Customer data and all serving stay in Greek data centres. US APIs are never in the serving path. Frontier models are used only as judges and teachers in evaluation.
Each is an open-weight base model with our own fine-tuned adapter. Every one ships with an AI Act (Regulation (EU) 2024/1689) dossier and a model card.
Model roles. Iris: the instant model, for Greek contact-centre text and call summaries. Metis: the thinking model, for document Q&A and agents that call tools. Clio: speech to text. Calliope: voice. Each is an open-weight base model with our own fine-tuned adapter; the base is named on every card. Llama and Qwen stay available as routing alternatives.
Named after the swift messenger. Greek contact-centre text and call summaries, answered at once.
Named after the goddess of counsel. Document Q&A on 12B; agents that understand the caller and call tools on 26B.
Named after the muse who records what was said. Streaming Greek transcription for telephone audio.
Named after the muse with the beautiful voice. One consistent Greek voice for every call.
* Measured on the Greek base model before the voice adapter was merged. A merged adapter does not change the size of the model.
Other open models, routed in where a task or a licence favours them.
FP8 with prefix caching. Never INT4 on Hopper GPUs: we measured it 8× slower.
An OpenAI-compatible API on GPUs in Greek data centres. Price your workload next to frontier APIs, then create an early-access account. An administrator reviews every account and grants access.
Training and serving are kept separate, and every serving site is in Greece.
What runs where today: Contact-centre workloads on dedicated capacity in Athens data centres. AI Employee on EU cloud until its 2027 migration to our Greek core. Training and evaluation at Greek sites and through PHAROS AI-Factory access; benchmarks on rented GPUs use non-customer data only.
Dedicated GPU capacity in Athens data centres operated by Greek providers.
Production serving, contracted per customerGreece's national AI factory at Lavrion, for fine-tuning and large evaluation runs through AI-Factory access.
Training and evaluation · subject to access grantOwned H200 nodes in Athens colocation as capacity insurance.
Planned 2027Tell us the calls, documents or questions you want handled. We measure it on your data before you commit.