JOB-1545
An applied-AI company strips personal and sensitive details out of enterprise data so that it can be used safely while keeping its structure and usefulness. Because the data spans text, documents, tables, images and vendor-specific formats, the team needs an ML engineer who can make the models behind this process measurably better in production.
Your focus
- Take ownership of a prioritized slice of the pipeline, such as named-entity recognition, entity resolution, structured extraction, document understanding or semantic review
- Turn observed model failures into a ranked roadmap of improvements
- Create active-learning loops that draw on automated model passes, LLM-assisted reviewing, clustering and confidence signals to choose which samples deserve hand-labeling
- Assemble benchmarks and datasets that are truly representative, tune thresholds and weigh precision, recall and data-utility tradeoffs
- Choose between heuristics, traditional ML, fine-tuned models, embeddings or LLMs depending on evidence
- Ship with reproducible artifacts, instrumentation and careful rollouts, while managing inference cost and latency
Profile
- Three to eight years as an ML or software engineer, with models serving live users
- A record of delivering precision and recall gains through error analysis, new labels and retraining
- Command of F1, calibration, class imbalance, noisy labels and distribution shift
- NLP experience such as NER, information extraction, document processing or de-identification
- Strong Python skills and comfort working inside production data pipelines
- High agency and daily use of AI engineering tools
On-site in New York, NY. Visa transfers are considered.
