AI Backend Engineer (Python, Agentic AI)

Що робити

Design and evolve production AI architecture covering agent orchestration, routing, tool calling, RAG, memory and structured outputs;
Select and integrate LLM, speech (STT/TTS) and vision models from different providers, balancing quality, latency, cost and EU data residency;
Define the data architecture for conversational, nutrition and wearable data to support personalization and a digital twin;
Translate product requirements into technical solutions and document key architecture decisions together with product owners;
Develop and extend the Python-based AI backend, including voice pipelines with real-time latency requirements;

6+ years of experience in software or ML engineering, with at least 2 years of hands-on experience building and running LLM-based systems in production;
Strong proficiency in Python, with solid SQL and data modeling skills;
Hands-on experience designing and implementing agentic systems using orchestration frameworks such as LangGraph or similar, including tool calling, structured outputs, RAG, and long-term memory;
Experience working with multiple LLM providers and model gateways, as well as prompt engineering and context engineering at scale;
Practical LLMOps experience covering evaluation frameworks, observability, and optimization of cost and latency;

Experience with real-time voice AI, including streaming STT/TTS, speech-to-speech models, and latency optimization;
Experience working with health, fitness, or wearable data, including time-series data and APIs such as HealthKit / Health Connect;
Experience with personalization, recommendation systems, A/B testing, and statistics;
Experience with vision models for food recognition;
Experience in regulated environments such as GD