Enjenyè IA (Pòs-Edisyon MT)
Sou pòs la
Anons ki anba a parèt an anglè, egzakteman jan ekip rekritman nou an te ekri l.
Verdacert provides USCIS-certified translations for immigration filings, court proceedings, university applications, and medical documentation, with a focus on Arabic, Farsi, Dari, Pashto, Urdu, Turkish, and other MENA and world languages. Our pipeline pairs a machine translation engine with native-speaker reviewers who edit and sign off on every document before delivery. We're looking for an AI Engineer to own and improve the machine-translation post-editing layer that sits at the center of this process, working closely with our reviewer network to raise draft quality, cut editing time, and keep turnaround inside our 14-to-48-hour delivery windows.
What you'll do
- Build and maintain machine translation post-editing workflows that feed our human review pipeline
- Evaluate and fine-tune MT models and prompts for accuracy across multiple language pairs, with attention to regional dialects and document conventions
- Design automated quality metrics (edit distance, reviewer correction rates, error taxonomies) to track MT output quality over time
- Work with native-speaker reviewers to collect structured feedback and translate it into model or pipeline improvements
- Build tooling that preserves document layout and formatting through the translation and certification process
- Monitor production translation quality and turnaround, and troubleshoot regressions or model drift
What we're looking for
- Experience building or fine-tuning machine translation systems, or applying large language models to translation tasks
- Familiarity with post-editing workflows and quality metrics used in professional or MT-assisted translation
- Solid software engineering skills, comfortable building production pipelines rather than one-off notebooks
- Experience working with human reviewers or subject-matter experts to improve model output
- Understanding of the tradeoffs between automation and human review in high-stakes, compliance-sensitive documents
- Strong written communication for documenting model behavior and quality issues to non-technical stakeholders
Nice to have
- Experience with low-resource or non-Latin-script languages (Arabic, Farsi, Dari, Pashto, Urdu)
- Background in localization, computational linguistics, or translation technology
- Familiarity with document layout preservation (OCR, PDF structure extraction) alongside text translation
Kijan pou poze kandidati
Kandidati pou pòs sa a pase sou Landearly, sèvis swivi kandida nou itilize a. Sa pran kèk minit; Landearly kenbe kandidati ou a epi ekip rekritman nou an revize li, jan sa dekri nan seksyon «Moun k ap poze kandidati» nan Politik Konfidansyalite nou an.
Sa a se yon pòs ki baze Ozetazini, konplètman a distans. Kandida yo dwe gen otorizasyon pou travay Ozetazini; kounye a nou pa fè sponsò viza epi nou pa anplwaye moun atravè yon anplwayè anrejistre aletranje.
Verdacert LLC se yon anplwayè ki bay chans egal. Nou konsidere chak kandidati san gade ras, koulè, relijyon, sèks, oryantasyon seksyèl, idantite jan, orijin nasyonal, laj, andikap, estati veteran, oswa nenpòt lòt karakteristik lalwa pwoteje. Si ou bezwen yon akomodasyon pou poze kandidati, voye yon imèl bay careers@verdacert.com.