For enterprise media too confidential for ChatGPT, DeepL, or a freelancer's laptop


Terminology is your asset. Preserve it precisely.

Vakyya is the UK and European pipeline for confidential internal media.

Transcribed, translated, captioned, and editorially enriched — without leaving your jurisdiction.

UK & EU hosted Signed DPA ICO registered UK GDPR & PECR

What we do


You upload the media. We return the transcript, the translations, and the captions.

With your terminology preserved, hosted in your jurisdiction throughout.

No more weekend workarounds. No uncontrolled files. No undocumented processing paths.

What makes Vakyya different


Three things generic AI tools and freelancers cannot give you.

What it isn't


Vakyya is not live captioning. Not a meeting assistant. Not a generic translation tool. Not a translation management system.

It is the managed workflow for confidential internal media that needs controlled processing, terminology consistency, and an audit trail.

Who it's for


Regulated UK and European organisations with multilingual workforces and recurring confidential media.

Financial services. Pharma services. Regulated training. Professional services.

Leadership briefings. Town halls. Compliance training. Policy updates. Medical affairs content. Specialist interviews.

Teams where "we translated it ourselves over the weekend" is no longer a sustainable answer.

Inside the pipeline


Six stages. One governed pipeline. Your terminology held across every one.

Internal media goes in. Captioned, translated, searchable and audit-ready content comes out. The pipeline is built for all the things that cannot drift: your product names, regulated terms, brand vocabulary, approved phrasing and the evidence trail behind every output.

  1. 01

    Ingest

    Upload media. Source files stay in UK, EU or Swiss storage from the moment of upload.

  2. 02

    Transcribe

    Speech-to-text with speaker diarisation. Source-language transcripts remain reviewable before translation begins.

  3. 03

    Glossary lock

    Customer terminology is applied before translation. Protected names, regulated terms and approved phrasing are pinned.

  4. 04

    Translate

    Output in target locales. Protected terms are preserved and substitutions are logged.

  5. 05

    Caption and package

    Caption files, transcript exports and media-ready assets are prepared for delivery.

  6. 06

    Enterprise tier

    Editorial enrichment

    Summaries, chapters, QA flags, glossary insights, metadata and audit-ready delivery packs, generated pre-handoff.

How it works


The architecture procurement keeps asking about.

Vakyya is a managed pipeline on Google Cloud Platform, hosted in London, Frankfurt and Zurich regions, with glossary enforcement, audit logging and reproducible runs built-in at the architecture layer rather than bolted-on at the API.

Vakyya reference architecture Three operationally independent horizontal processing pipelines stacked vertically: the UK region (top) hosted on GCP London (europe-west2), the EU region (middle) hosted on GCP Frankfurt (europe-west3), and the Swiss region (bottom) hosted on GCP Zurich (europe-west6). Customer uploads are routed by jurisdiction at the point of upload — UK customer content flows into the UK region, EU customer content flows into the EU region, and Swiss customer content flows into the Swiss region. Each region contains an identical pipeline of six stages flowing left to right: source storage, transcription, glossary engine, translation, caption plus package, and editorial enrichment. Within each region, the translation stage stacks two configurable options vertically: a managed EU translation processor as the default path on top, and in-region inference as the enterprise tier option beneath. The editorial enrichment stage is marked as an enterprise tier stage. Optional encrypted disaster-recovery replication may connect source storage and the editorial enrichment stage between adjacent regions for enterprise tier customers; transcription, glossary engine, translation and editorial enrichment inference never cross regions. VAKYYA PREMIUM REFERENCE ARCHITECTURE Customer upload UK customer content EU customer content Swiss customer content UK REGION · GCP LONDON (EUROPE-WEST2) Source storage Transcription Glossary engine default Managed EU translation processor enterprise tier In-region inference Caption + package Editorial enrichment enterprise tier EU REGION · GCP FRANKFURT (EUROPE-WEST3) Source storage Transcription Glossary engine default Managed EU translation processor enterprise tier In-region inference Caption + package Editorial enrichment enterprise tier SWISS REGION · GCP ZURICH (EUROPE-WEST6) Source storage Transcription Glossary engine default Managed EU translation processor enterprise tier In-region inference Caption + package Editorial enrichment enterprise tier Opt-in encrypted DR replication Enterprise tier only Opt-in encrypted DR replication Enterprise tier only IN-REGION FLOW OPT-IN DR REPLICATION DEFAULT PATH ENTERPRISE TIER OPTION

Each region is operationally independent. Customer content is routed by jurisdiction at upload and remains in its region throughout the pipeline. Cross-region replication is available on enterprise tier for disaster recovery, scoped to source and delivered artefacts only. Inference never crosses regions.

  • Data residency. Source files, intermediate artefacts and outputs all remain in UK, EU or Swiss storage. No US-hosted inference, no cross-border data transfer, no unmanaged public-model submission.

  • Glossary enforcement. Per-customer term lists are applied at the model layer using span-based placeholder substitution. The terms you protect are the terms that ship.

  • Audit trail. Every transcription run, translation decision and glossary substitution is logged. Reproducible runs by job ID. Exportable for legal review.

  • Built for the boring bits. Signed DPA, sub-processor list, retention policy, ICO registration, UK GDPR and PECR posture. The compliance binder writes itself.


Engineering provenance

Vakyya's glossary preservation layer was developed alongside a University of Cambridge postgraduate programme in Data Science with Machine Learning and AI. The methodology was developed against one of the hardest terminology problems in NLP: preserving a 216,000+ term specialist vocabulary in a low-resource language with no margin for error.

Your product names and regulated terminology are an easier problem than the one this system was designed for.

Compliance posture


The answers your procurement team needs, all written down.

Vakyya is built for buyers who are obliged to evidence their data handling. Every claim below is documented and available under NDA on request.

Hosting
Google Cloud Platform, London (europe-west2), Frankfurt (europe-west3) and Zurich (europe-west6) regions only.
Data residency
Source content, transcripts, translations and audit logs remain in UK, EU or Swiss storage throughout the pipeline.
Sub-processors
Published list, version-controlled. Notification on change.
Data Processing Agreement
Signed DPA available on request. UK GDPR and EU GDPR aligned.
Retention
Configurable per contract. Default: source content deleted 30 days after delivery; audit logs retained 12 months.
Encryption
AES-256 at rest, TLS 1.3 in transit. Customer-managed encryption keys available on enterprise tier.
Regulatory
ICO registered. Aligned to UK GDPR and PECR obligations, with signed DPA, retention controls, and disclosed sub-processors.
Certifications
ISO 27001 and SOC 2 engagement timeline available on request.

Detailed compliance documentation available under NDA. Contact us for the full pack.

Next step


Twenty minutes. No pitch deck.

A short call to understand what you're translating, where the media is allowed to go, and what terminology cannot drift. Bring one real workflow if you have it and we'll tell you whether Vakyya fits. If it doesn't, we'll be completely honest with you.