Docs / 02-requirements/003-bilingual-greek-english-llm

REQ-003 Greek + English language support across the AI platform

The LLM environment, RAG, and all assistants must work well in both Greek and English — queries, documents, and outputs.

type: requirement updated: 2026-07-03 owner: IT lead

REQ-003 — Greek + English language support across the AI platform

Need (what is being asked)

The entire AI environment must support Greek and English: chat, document understanding, retrieval (including Greek query → English document and vice versa), transcription of Greek meetings, and generation of client-facing English output.

Context (why — what problem it solves)

Internal work is largely Greek; shipping clients and suppliers work in English; tenders and ISO documentation are bilingual. Weak Greek performance would kill adoption.

Current state (how it is done today)

Manual bilingual work by staff; no tooling.

Success criteria

Greek eval set (built from real company documents, ~M4) meets the quality bar before platform GA; embedding model retrieval works cross-lingually; Whisper transcription of Greek meetings is usable.

Model choices: llm-serving-stack. Risk: Greek-language LLM performance (risk-register).

History

2026-07-03 — captured (source: CEO transformation brief)