Lead Machine Learning Engineer (Personalization & AI Models)

About the role

We are looking for a Lead Machine Learning Engineer to design, build, and optimise personalisation models for a large‑scale mobile app. The role focuses on multi‑vector representations, real‑time inference, and integration with AWS Personalise, PGVector and RudderStack.

Key responsibilities

• Develop and optimise embedding models using sentence‑transformer techniques for user profiling.
• Implement KNN‑based recommendation systems for real‑time content scoring and ranking.
• Use AWS Personalise SDK to train, deploy and continuously adapt models to user behaviour.
• Integrate embeddings with Databricks, RudderStack and AWS services for live profile updates.
• Fine‑tune models to improve revenue prediction, engagement and audience segmentation.
• Optimise PGVector and Redis for fast vector look‑ups and caching.
• Collaborate with data engineers to build robust ML pipelines for training, validation and deployment.

Required profile

• Strong experience in machine learning, deep learning and AI‑driven personalisation.
• Proficiency in Python and modern DL frameworks (PyTorch, TensorFlow).
• Hands‑on experience with vector‑based search, KNN, PGVector and Redis.
• Familiarity with AWS Personalise SDK, SageMaker or similar training pipelines.
• Experience with Databricks, Delta Lake and Apache Spark for large‑scale model training.
• Understanding of real‑time personalisation, A/B testing and event‑driven architectures.
• Fluent English communication (B2+).

Required skills

• Python
• PyTorch
• TensorFlow
• Sentence‑transformer models
• KNN algorithms
• PGVector
• Redis
• AWS Personalise SDK
• SageMaker
• Databricks
• Delta Lake
• Apache Spark

What we offer

• Remote work from anywhere in Ukraine.
• Flexible schedule and modern coworking spaces.
• Access to diverse projects across insurance, e‑commerce, healthcare, finance, media and AI.
• Opportunities for relocation within EPAM.
• Professional growth programmes, mentorship and regular feedback.