Senior AI Platform Architect & Infra Lead

EPAM Systems; United States; Remote; USD 180,000 - 240,000; Full time

Job summary

EPAM Systems is seeking a Principal AI Solution Engineer to join an engineering organization building a sovereign, multi-tenant agentic AI platform with Arabic and English support. Squads are small and decisions are made in the room, requiring credibility across applied AI, AI infrastructure and backend platform.

You will own architecture across three layers, lead GPU serving (vLLM, TGI or TensorRT-LLM), define RAG and agent infrastructure, and ensure robust, event-driven services with clear

Qualifications

• 5+ years of hands-on engineering experience across AI, infrastructure and backend platforms.
• At least 1 year of leadership experience.
• Proficiency in Python production stack (FastAPI, Pydantic, asyncio).
• Experience with production RAG end‑to‑end: ingestion, vector search, reranking and grounding evaluation.

Responsibilities

• Define architecture spanning squads; document decisions and defend them.
• Build and tune GPU-based serving (vLLM, TGI or TensorRT-LLM) for latency, throughput and cost per token.
• Own the reference RAG and agent architecture used by other teams.
• Define service boundaries, API contracts, event-driven infrastructure and the SLO work that prevents incidents.
• Develop offline and online evaluation and regression detection to evolve models safely.

Skills

Leadership

Distributed systems

Architectural decision making

English communication

AI platform experience

Tools

Python

FastAPI

Pydantic

asyncio

vLLM

TGI

TensorRT-LLM

LangGraph

LangChain

OpenAI Agent SDK

Kubernetes on Azure

PostgreSQL

CLAUDE.md

Job description

EPAM Systems is seeking a Principal AI Solution Engineer to join an engineering organization building a sovereign, multi-tenant agentic AI platform with Arabic and English support. Squads are small and decisions are made in the room, requiring credibility across applied AI, AI infrastructure and backend platform.

You will own architecture across three layers, lead GPU serving (vLLM, TGI or TensorRT-LLM), define RAG and agent infrastructure, and ensure robust, event-driven services with clear