Senior Applied Scientist / MLOps Engineer (Privacy AI)

Senior Applied Scientist/MLOps Engineer for Privacy AI - data de-identification. Full-time, remote in Russia, contract with individual entrepreneur, 1200-1500 RUB/hour.

Develop data classifier for database fields (Tabular ML):- Improve existing RandomForest model to identify personal data in DB fields based on metadata and values, including new types (VIN, specific identifiers) and mixed data (names, addresses, phones).- Optimize model for high-speed CPU inference to scan multiple TB databases.Build NER model for detecting personal data in unstructured text (NLP/NER):- Design architecture and train NER model for unstructured text.- Process various file formats (TXT, DOCX, PDF, logs).- Implement pipelines for cleaning and normalizing noisy data.

Experience and leadership:- 3+ years of experience developing and deploying ML models to production.- Full cycle experience: from research and problem definition to quality monitoring and automatic retraining setup.- Leadership: ready to lead the direction, decompose business tasks into technical tasks, and take responsibility for results.Technical stack (Python and ML):- Strong Python skills (pandas, numpy, scikit-learn).- Deep understanding of classical ML algorithms (decision trees, ensembles, Random Forest), ability to interpret results (Feature Importance) and tune hyperparameters.NLP:- Practical experience with Hugging Face Transformers and/or SpaCy.- Mandatory experience training NER models.Data Engineering:- Experience building pipelines for unstructured data (PDF, texts).- Feature Engineering for tabular data.Performance and infrastructure (MLOps):- Experience optimizing model inference for CPU.- Experience with large data volumes (full scan of databases from several TB).- Experience deploying models via API (FastAPI), containerization (Docker).- MLOps: versioning tools (MLflow, DVC), understanding CI/CD processes, experience setting up model quality monitoring (Model Drift, Quality Gates) and automatic retraining.Nice to have:- Experience with classifiers for DB data profiling.- Experience recognizing specific data types (VIN, card numbers, crypto addresses).- Experience in personal data protection, data de-identification, or anti-fraud systems.

Opportunity to influence the architecture of a complex AI product from scratch.Work with real Big Data tasks (TB+).Modern stack and no legacy code.Contract with individual entrepreneur.Payment: from 1200 to 1500 RUB/hour.Full-time, long-term.