platform engineer

EPAM develops enterprise software products, open source solutions, and accelerators. Its Enterprise Agent Development Platform enables engineering teams to define, orchestrate, deploy, and observe AI agents at scale.

• Develop and launch scalable conversational AI systems across chat, voice, email, and APIs;
• Architect microservices and asynchronous systems for AI applications with high availability and performance;
• Use Python to implement orchestration and decision logic for tools, workflows, and system interactions;
• Design agentic evaluation frameworks for multi-step workflows, tool-calling accuracy, and session performance;
• Integrate automated evaluation gates into CI/CD pipelines;
• Establish automated evaluation protocols using LLM-as-a-judge patterns;
• Optimize AI models through fine-tuning and LoRA integration;
• Make data-driven decisions between fine-tuning and RAG approaches;
• Work with data specialists to design pipelines for collecting, cleaning, and structuring information for AI agents;
• Design CI/CD pipelines, container standards, and cloud strategies for infrastructure migrations, cost optimization, and high availability.

• Strong Python skills for integrating AI models, automation, and orchestration logic;
• 1–2 Years of hands-on experience with Generative AI and large language models;
• Practical machine learning engineering experience, including data preparation, cloud vector databases, and model training or optimization;
• Hands-on experience designing evaluation frameworks for complex multi-agent systems, including reasoning traces, tool selection, and agentic workflows;
• Proven experience building automated testing and evaluation gates within CI/CD pipelines for AI models;
• Strong backend and frontend architecture experience;
• Experience with AWS, GCP, or Azure, including container orchestration and CI/CD pipelines;
• Experience with agentic frameworks such as LangChain and LangGraph and with multi-agent system design;
• Nice to have: TensorFlow, PyTorch, Scikit-learn.

• No conditions specified