Arutiun Dzhilavyan

I’m a Machine Learning Engineer with 5+ years of experience designing scalable ML architectures and leading end-to-end deployment of production-grade machine learning systems.

At Provectus, I build GenAI on AWS Bedrock—multi-agent pipelines, RAG systems, and document intelligence platforms—then operationalize them with evaluation/refinement loops, cost-controlled inference, and secure guardrails (PII masking and prompt injection protection). I also implemented prompt management with LangSmith (LangGraph state machine) and deployed LLM microservices using FastAPI and AWS serverless components.

Previously, at AraratBank, I led Probability of Default (PD) models into enterprise IFRS 9 production pipelines and built automated ECL reporting using Python, Airflow, and CI/CD. Earlier roles covered NLP/semantic search with transformer architectures, scalable ETL and retraining workflows (Airflow, Docker, AWS/GCP), and production ML microservices—plus monitoring and retraining strategies that keep models reliable over time.