Senior AI Translation Engineer (NLP)
Mission
As a AI Translation Engineer, you will be at the forefront of advancing our platform’s translation capabilities. Leveraging your expertise in machine learning including applied research, fine tuning, benchmarking, and inference optimization, you will deliver high-performance solutions that power critical features for our growing client base.
Requirements
Evaluation engineering: Translates product features into measurable user use cases; selects a small set of reliable offline metrics; builds benchmarks and runs slice analysis across languages and domains; designs rubrics and reference-less MT evaluation for ambiguous cases; defines online metrics and guardrails, runs A/B tests and staged rollouts, and validates that offline gains translate into production impact while spotting metric loopholes and regressions
Data fluency: Owns end-to-end dataset pipelines: multi-source ingestion → cleaning/validation → scalable multilingual datasets, with versioning and auditability
Classical NLP foundations: Understands MT/search text normalization and core matching approaches (exact/fuzzy, typo tolerance, multi-word + overlap handling, tries/Aho–Corasick/inverted index basics), and applies them to improve preprocessing and retrieval
LLM operations: Operates external LLM providers end-to-end: designs prompts for structured generation/evaluation; builds automated filters/validators; generates synthetic data at scale using self-consistency and multi-provider “consilium”; selects and routes across providers based on quality–latency–cost; and uses self-hosted models when helpful for quality analysis/judging and provider fine-tuning data prep
Practical production engineering: Can write lightweight services/jobs when needed and ship them with structured logs, monitoring/alerts, reliability basics (retries, idempotency, fallbacks), and scalability considerations; collaborates effectively with Dev/Platform
Key Responsibilities
Develop and improve benchmarks for translation problems
Work with NER, word alignment, automatic translation quality scoring system, and other problems
Set up infrastructure for self-hosted solutions
Our technologies
ML/Backend: Python, C#, PyTorch, Transformers
DB: MongoDB, PostgreSQL, Elasticsearch
Messaging: Apache Kafka
Cloud: Amazon AWS
Monitoring: ELK, Prometheus, Grafana