Overview
What this challenge is about.
Pick an open MT base (NLLB-200 or a strong open M2M model). Build a parallel corpus of around 8,000 sentence pairs from the company's bilingual safety standards. Fine-tune on the corpus, paying special attention to terminology preservation and cross-reference handling (e.g., '§4.2.3' must survive). Evaluate on a 300-sentence held-out test with both automated metrics (BLEU, COMET) and a human-judged terminology-accuracy score. Compare against generic MT. Deliver the model, terminology glossary, and a 4-page integration memo.
The Brief
What you'll do, and what you'll demonstrate.
Adapt an open MT system to automotive-safety German-English with measurable terminology accuracy beyond generic MT.
Earning criteria — what you'll demonstrate
- Build a parallel corpus from real bilingual documents
- Fine-tune neural MT for domain terminology
- Evaluate MT with both automated and human-judged metrics
- Preserve structured tokens (cross-refs, abbreviations) through MT
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
NLP Engineer
Domain-adapting MT systems and proving terminology accuracy is the work NLP engineers do at any company with multilingual technical documentation.
This challenge sharpens
- machine-translation
- domain-adaptation
- terminology-management
Applied AI Scientist
Combining automated and human evaluation, and reasoning about constrained decoding, is core applied-AI-scientist work in MT and structured-text NLP.
This challenge sharpens
- machine-translation
- evaluation
- domain-adaptation
Machine Learning Engineer
Shipping a fine-tuned MT model and the inference pipeline to an engineering team is the MLE work that vertical AI companies need.
This challenge sharpens
- transformers
- pytorch
- machine-translation