ai engineer

EPAM provides digital engineering, cloud, and AI-enabled transformation services, as well as business and experience consulting, to global enterprises.

• Design agent orchestration including graph/state, conditional routing, tool calling, memory, and checkpointing;
• Build production RAG end-to-end covering chunking, embeddings, vector stores, hybrid retrieval, reranking, caching, and grounded synthesis;
• Own Python and FastAPI services including async, SSE streaming, session handling, and structured error contracts;
• Instrument systems with tracing and evaluation harnesses for accuracy, cost, and regression;
• Ship services on Docker and Kubernetes via CI/CD with test, eval, and canary gates;
• Drive LLM cost engineering including model routing, prompt optimization, caching, token accounting, and build-vs-buy decisions;
• Apply GenAI safety and governance such as hallucination control, prompt-injection defense, PII handling, and human-in-the-loop processes;
• Partner with data engineering on semantic layers and pipelines.

• 5+ Years in software engineering, with 2+ years shipping production LLM or agentic systems;
• Proficiency in Python and FastAPI;
• Production expertise in LangChain and LangGraph or equivalent stacks;
• Background in production RAG including embeddings, chunking, and hybrid retrieval;
• Skills in vector databases such as Pinecone, Weaviate, pgvector, OpenSearch, or Databricks Vector Search;
• Knowledge of at least one major LLM provider in production with model selection and routing trade-offs;
• Competency in Kubernetes and Docker in production environments;
• Expertise in cloud engineering on AWS;
• Familiarity with observability and tracing tools, evaluation harnesses, and latency/cost ownership;
• Capability to build CI/CD for AI systems with test/eval gates;
• Strong written and spoken English (B2 level);
• Nice to have: Databricks depth, experience with LLM fine-tuning, understanding of MCP servers and tool integration, qualifications in GenAI governance and FinOps, background in classical ML or DL.

• No conditions specified