LLM Tech Lead at Raiffeisenbank

Lead architect for AI agents infrastructure; designed and deployed end-to-end stack including Milvus vector search, Langfuse observability, and a secure Kubernetes code-execution sandbox. Implemented PEFT methods (LoRA, QLoRA) for fine-tuning VLLM; built RAG-based search ranking increasing relevance by 30%. Developed chatbots for support staff covering 15% of requests. Created transcription, diarization and summarization service for support calls with ASR/LLM; analyzed topics, intents and emotions to reduce Mean Handling Time by 20% and improve First Call Resolution by 15%. Deployed microservices in Kubernetes for high availability and scalability.

ML Engineer at OZON.ru

Worked in the matching team; developed and optimized product matching pipelines; trained and deployed transformer-based text models for search and ranking; applied dimensionality reduction to improve efficiency. Built and deployed classification models for deduplication. Created Python+gRPC microservices for real-time matching; implemented metric monitoring in Grafana. Led multiple A/B tests and explored automatic retraining of classification models.