We are an international developer of IT products for medicine and related services. Our products are used by more than 5,000 doctors and more than 1,000 clinics in Europe.
The goal is to attract the largest possible audience to the main projects - clinics and doctors up to 3 million. As well as expanding unique traffic to 10 million per day until 2025. We are looking for AI Engineer / AI Architect with deep knowledge of Fundamental Mathematics, Probability Theory and Bayesian Statistics, Graph Algorithms, Mathematics of Transformers, Hybrid AI Approaches, Uncertainty Assessment.
Must-have skills are: 1. ML and Data Science (Core): - Expert in NLP & LLM: Deep understanding of Transformers architecture (BERT, RoBERTa, GPT, T5, Llama/Mistral). Experience with Large Language Models (LLM) – not prompt engineering, but fine-tuning (PEFT, LoRA) of open-source models on specific datasets. - RAG (Retrieval-Augmented Generation): Experience building RAG systems is required. The AI must be able to search for information in different data sources. - Named Entity Recognition (NER) & Information Extraction: Ability to extract specific entities from unstructured text. - Tabular Data & Classic ML: Confident use of classical algorithms (XGBoost, CatBoost, LightGBM) for analyzing structured data and building probability scoring models. - Frameworks: PyTorch (preferred) or TensorFlow/Keras. Proficiency in HuggingFace, Scikit-learn, Pandas, and NumPy.
2. MLOps: - Experience wrapping models in APIs (FastAPI, Flask) with minimal latency.
- Docker knowledge and experience working with local GPU instances are required. - Experience working with vector databases (Pinecone, Milvus, Qdrant, ChromaDB) for quickly searching for similar cases or working with knowledge bases.
- ML Pipelines: Experience building model training and retraining pipelines (Airflow, MLflow, DVC).
3. MedTech and Security: - Practical working experience with sensitive personal data and GDPR/DSGVO requirements.
- It is important to be able to deploy and optimize local models (Llama 3, Mistral, Mixtral) within a closed company environment to ensure data privacy. - Skills in quantization and distillation of models to speed up work without significant loss of quality.
Nice-to-have skills: - Experience working specifically with medical data (EHR, EMR; HL7, FHIR, ICD-10/11 standards).
- Knowledge of Interpretability/Explainable AI (XAI) methods. The end user should understand why the AI suggested a particular option (SHAP, LIME).
- Experience with multimodal models.
What we offer: - Paid holiday for 28 working days. - Transparent incentive system.
