What Prompsit Language Engineering does

Prompsit Language Engineering, S.L. is a Spanish private company founded in 2006 as a university spin-off from the Transducens research group at Universitat d'Alacant, headquartered at the Miguel Hernández University Science Park in Elche. The firm specializes in multilingual language technology, combining the curation and enrichment of multilingual corpora with the training and deployment of domain-adapted machine translation and language models. Its technical stack integrates rule-based MT (Apertium), neural MT (Mozilla & OPUS-MT via CTranslate2), language variant conversion (AltLang), and 15+ open-source corpus-processing tools (Bitextor, Bicleaner, Bifixer, OpusCleaner, OpusTrainer, HPLT Analytics, KEOPS, Corset, Monotextor, Biroamer, Term Gallery).

The commercial offering is built around three product lines: Smart Datasets (collection, cleaning, alignment, enrichment, and synthetic data generation for multilingual corpora), Smart Models (fine-tuning and deployment of MT/LLM models for specialized domains), and the Prompsit API (a privacy-first, flat-subscription translation and evaluation service). Smart Datasets and Smart Models are delivered primarily as professional services or managed engagements, with on-premises or private VPC deployment targeted at regulated customers in legal, medical, financial, and public-sector organizations requiring GDPR and EU AI Act compliance. The Prompsit API is sold on an annual flat subscription, positioned against per-token hyperscaler pricing models.

Customer concentration is split between named enterprise clients (Adobe, Autodesk, Tripadvisor, Smartling, Across, Reverso, LinguaServe) and European public-sector and nonprofit accounts (Generalitat Valenciana, La Caixa, Repsol, GenCat, Translators without Borders, Translation Commons, Universitat d'Alacant). Go-to-market combines direct enterprise sales for regulated engagements with self-serve API access and research-driven demand generation via 30+ academic publications, 1,800+ citations, and active participation in EU-funded projects including HPLT, OpenEuroLLM, MaCoCu, ParaCrawl, MultiTraiNMT, and EuroPat. Revenue figures are not publicly disclosed; at 1-10 employees, the company operates as a boutique specialist rather than a scaled software vendor.

Name

Prompsit Language Engineering

Legal name

Prompsit Language Engineering, S.L.

Website

https://prompsit.com

Company type

Private

Founded year

2006

Operating status

Operating

Headcount range

1–10 employees

Short description

Prompsit Language Engineering curates multilingual datasets and trains domain-adapted machine translation and language models for European enterprises, public-sector bodies, and regulated industries needing GDPR- and EU AI Act-compliant language technology.

Ownership category

akta.pro rank

Product category

Machine Translation Software

NAICS

Software Publishers (513210), Custom Computer Programming Services (541511), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518)

SIC

Services-Computer Processing & Data Preparation (7374), Services-Prepackaged Software (7372)

akta.pro primary industry

Machine Translation & Multilingual NLP (HDAAADAD)

akta.pro secondary industry

Localization, Translation & Multilingual Publishing Tools (EDAFAHAJ)

Keywords

• Machine translation services
• Multilingual data curation
• Language technology APIs
• Neural translation models
• Open-source NLP tools

Headquarters

HQ city

Elche

HQ country

Spain

HQ region

Europe

Offices1 record

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel

Revenue model

1. Prompsit API Subscriptions: Flat subscription-based pricing for API access providing translation, evaluation, scoring, and annotation services. Rate and volume limits prevent abuse rather than meter cost.
2. Custom Model Training & Fine-tuning: Domain-specific machine translation and language model customization services for regulated industries including legal, medical, technical, and financial sectors with enterprise terminology requirements.
3. Smart Datasets Services: Data curation, collection, cleaning, alignment, and enrichment services for multilingual corpora tailored to client domain and language requirements.
4. Model Deployment Services: On-premises or private cloud VPC deployment of MT and LLMs with security controls, version tracking, and auditing for regulated sectors.
5. Open-Source Tools & Community: Open-source software contributions (Apertium, Bicleaner, Bifixer, Bitextor, etc.) that build community adoption and drive enterprise adoption of commercial services.

Pricing tiers

Go-to-market motion2 records

Distribution channels4 records

Marketing channels5 records

Core offering

Prompsit curates and transforms multilingual data into language technology services. Its core commercial offerings are Smart Datasets (multilingual corpus collection, cleaning, alignment, and synthetic data generation) and Smart Models (MT/LLM fine-tuning, evaluation, and on-premises or cloud deployment for regulated industries), delivered alongside the Prompsit API for translation, evaluation, scoring, and annotation across 200+ languages including low-resource variants.

Product overview

Prompsit Language Engineering offers a platform-plus-services architecture centered on AI-powered language technology. The core offerings comprise Smart Datasets (data curation, cleaning, alignment, and enrichment for multilingual corpora) and Smart Models (fine-tuning, evaluation, and deployment of MT/LLM for specialized domains). The Prompsit API provides unified access to translation, evaluation, and annotation capabilities combining neural MT (Mozilla & OPUS-MT via CTranslate2), rule-based MT (Apertium), and language variant conversion (AltLang). The product portfolio includes 15+ open-source tools for corpus processing, cleaning, alignment, and evaluation, including Bicleaner, Bifixer, Bitextor, KEOPS, HPLT Analytics, OpusCleaner, OpusTrainer, and Apertium (a rule-based MT platform). Services target regulated industries (legal, medical, financial, public sector) requiring data privacy, EU AI Act/GDPR compliance, and on-premises deployment options.

Differentiator

Problem solved

Functional benefit

Brands

• Apertium: Free and open-source platform for rule-based machine translation, supporting dozens of languages.
• AltLang
• MutNMT
• Bicleaner
• Bifixer
• Bitextor
• KEOPS
• Corset
• HPLT Analytics
• Monotextor
• Biroamer
• Term Gallery
• Promut
• Prompsit API

Products and services

• Prompsit API Unified translation, evaluation, scoring, and annotation API combining Mozilla & OPUS-MT (CTranslate2 neural MT), Apertium (rule-based MT), and AltLang (language variant conversion) with flat-rate subscription pricing, configurable data retention, async document translation, and CLI-first developer workflows via npm.
• Smart Datasets Multilingual corpus curation and enrichment service covering data collection, cleaning, normalization, parallel alignment, analysis, and synthetic data generation tailored for domain adaptation and LLM training, with GDPR/EU AI Act alignment.
• Smart Models Machine translation and language model fine-tuning service for specialized domains including legal, medical, technical, and financial sectors. Includes quality auditing, on-premises or cloud deployment, secured development environments, and ongoing performance monitoring.
• Apertium Free and open-source platform for rule-based machine translation, supporting dozens of languages including regional and low-resource variants such as Catalan, Galician, Norwegian Nynorsk, and Valencian.
• AltLang Automatic language variety converter that adapts content between regional variants of English, Spanish, French, and Portuguese (e.g., en-US/en-GB, pt-BR/pt-PT, fr-CA/fr-FR, es-LA/es-ES) while preserving meaning and terminology.
• Bicleaner Open-source tool for detecting noisy sentences in parallel corpora using neural classifiers, used in dataset cleaning pipelines for machine translation training.
• Bifixer Open-source tool for parallel data cleaning, part of the dataset processing toolkit for normalizing and fixing alignment issues in bilingual text.
• Bitextor Open-source web-based application for harvesting parallel corpora from multilingual websites, used in Prompsit's data acquisition pipeline.
• HPLT Analytics Analytics tool for evaluating large-scale datasets, used in the HPLT project for auditing corpora containing billions of documents and segments.
• OpusCleaner Open-source data downloading, cleaning, and preprocessing toolkit for building MT and LLM training pipelines from various sources.
• OpusTrainer Open-source data scheduling and augmentation tool for building large-scale MT systems and LLMs, with deterministic data mixing, on-the-fly augmentation, and multilingual/terminology-aware model training.
• MutNMT Educational neural machine translation web application for learning NMT concepts, developed within the MultiTraiNMT Erasmus+ project.
• KEOPS Tool for manual evaluation of corpora for multilingual tasks, streamlining evaluation processes and supporting quality decision-making.

Quantifiable outcome

• MaCoCu and OSCAR achieve best intrinsic quality among web-crawled corpora for language model training
• +1 more outcomes

Named customers9 records

Segments5 records

Ideal customer profiles5 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration10 records

AI capability4 records

Feature10 records

Partnerships

Nine partnerships are on record, tiered flagship, core and minor.

• OpenEuroLLMflagshipStrategic or Co-development Partner · 1 January 2022Prompsit participates in OpenEuroLLM, a European initiative to promote open, transparent language models and strengthen European digital sovereignty. The project aims to ensure any European entity can adapt AI models to their needs while complying with data protection regulations.
• HPLT (High Performance Language Technologies)coreStrategic or Co-development Partner · 1 January 20223-year EU-funded project building massive multilingual datasets and models. Prompsit contributes expertise in corpus processing and curation, auditing corpora containing billions of documents using HPLT Analytics tools.
• MaCoCu (Massive Collection and Curation)coreStrategic or Co-development PartnerCEF-funded project building monolingual and parallel corpora for under-resourced European languages. Prompsit contributes web crawling and curation expertise with tools like Bitextor.
• ParaCrawlcoreStrategic or Co-development PartnerCEF-funded initiative for web-scale parallel corpora for EU languages. Prompsit contributed Bifixer and Bicleaner tools for cleaning noisy crawled parallel data.
• MultiTraiNMTminorStrategic or Co-development PartnerErasmus+ project developing open innovative syllabus in neural machine translation for language learners and translators. Prompsit contributed MutNMT educational platform.
• EuroPatminorStrategic or Co-development PartnerEuropean parallel corpus project with Prompsit involvement in corpus development and NLP tools.
• Transducens (Universitat d'Alacant)coreStrategic or Co-development PartnerPrompsit is a spin-off from the Transducens research group at Universitat d'Alacant, maintaining deep academic ties and collaborative research relationships.
• Mozilla & OPUS-MTcoreTechnology or IntegrationIntegration of Mozilla's OPUS-MT neural machine translation models via CTranslate2 into the Prompsit API for general translation across supported language pairs.
• Apertium CommunitycoreTechnology or IntegrationPrompsit maintains the Apertium open-source rule-based MT platform and contributes to its development. Apertium provides specialized coverage for regional European languages.

Scale indicators8 records

Recent moves7 records

Expansion highlights6 records

Direct peers

• DeepL: German neural MT provider offering a high-quality translation API and enterprise tier with on-premises deployment. Comparable to Prompsit in European origin, MT API business model, and emphasis on enterprise-grade quality and security.
• Unbabel: Lisbon-based AI translation platform combining neural MT with human post-editing, serving enterprise customer support and localization workflows. Closely comparable to Prompsit in European heritage, focus on translation quality, and enterprise go-to-market.
• Translated: Rome-based language services company operating its own MT platform (ModernMT) alongside professional translation services. Comparable to Prompsit in combining proprietary MT technology with enterprise localization offerings and serving multilingual content needs.
• LanguageWire: European language technology company offering translation management, MT integration, and enterprise localization services. Comparable to Prompsit in combining MT technology with enterprise localization workflows and European market focus.
• Lokalise: Cloud-based localization platform serving software companies with translation automation and developer workflows. Comparable to Prompsit in serving enterprise software localization needs and offering API-first translation tooling.
• Lengoo: Berlin-based custom neural MT platform providing domain-adapted enterprise translation with data security controls. Comparable to Prompsit in focusing on custom model training for regulated enterprises and offering private deployment options.

Broad incumbents

• RWS Group: UK-based global leader in language services, intellectual property, and regulated-industry localization with significant MT capability. Comparable to Prompsit in serving regulated enterprise localization buyers and combining MT technology with linguistic services.
• TransPerfect: World's largest language services provider offering GlobalLink translation management with integrated MT and on-premises deployment. Comparable to Prompsit in enterprise localization technology and regulated-industry deployment options.
• Welocalize: Global language services provider with proprietary MT and AI data services, serving regulated industries including legal, financial, and life sciences. Comparable to Prompsit in serving regulated verticals and offering custom MT plus data curation.

Emerging players

• Hugging Face: Open-source NLP model and dataset hub hosting many multilingual translation models including those built on OPUS-MT and Apertium-adjacent technologies. Comparable to Prompsit in open-source NLP ecosystem participation and multilingual model distribution.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks6 records

Key highlights7 records

Customer concentration

• GitHubgithub.com/apertium
• Xx.com/prompsit

Compliance2 records

Revenue estimate

Valuation estimate

Number of profiles

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Prompsit Language Engineering