machine learning engineer

Acba Bank OJSC operates in the finance, banking, and insurance industry.

• Design, develop, and deploy complex agentic workflows and automation ecosystems;
• Securely expose internal systems as tools via MCP and engineer stateful multi-agent systems;
• Optimize Time-To-First-Token and Tokens/sec for inference on in-house multi-GPU nodes;
• Architect systems capable of handling thousands of concurrent requests;
• Design fault-tolerant systems that gracefully handle the unpredictability of LLMs;
• Spearhead the evaluation and testing of agentic workflows;
• Mentor junior engineers through thoughtful code reviews and design feedback;
• Maintain up-to-date knowledge of related MLOps and data science topics and technologies;
• Build and optimize data pipelines, ETL processes, and model-serving frameworks;
• Model business requirements into structured software plans.

• Bachelor’s or Master’s degree in Computer Science, Engineering, or a related field;
• 3+ Years of experience building and operating scalable distributed and high-availability AI systems, with at least 2 projects shipped to production;
• Demonstrated experience building and deploying agentic systems, chatbots, or intelligent automation workflows;
• Exceptional proficiency in Python and FastAPI, with a strong understanding of OOP, software design patterns, and clean architecture;
• Experience with Docker and Kubernetes containerization;
• Experience with agent orchestration frameworks such as LangGraph and the MCP protocol;
• Experience with RAG architectures, vector databases such as FAISS, Qdrant, and Milvus, and semantic retrieval systems;
• Extensive experience with SQL, including T-SQL and Python’s SQLAlchemy toolkit, and NoSQL databases such as Redis;
• Strong engineering rigor, including commitment to TDD, automated testing strategies for ML models, and building highly observable AI systems;
• Evaluation-centric approach to building AI systems and deep understanding of classical and LLM metrics, including precision/recall, BLEU, ROUGE, faithfulness, and answer relevance;
• Expertise in MLOps frameworks such as MLflow and Kubeflow;
• Nice to have: On-premise LLM deployments, PEFT techniques including LoRA, QLoRA, Prefix Tuning, and FSDP, proficiency in C, C++, Rust, or Go, open-source contributions, well-documented learning journeys.

• No conditions specified