AI Engineer — Prague (Remote / Hybrid, Contract)

You will be responsible for designing and optimizing the multi-phase RAG pipeline at the core of the product — from vector search and LLM orchestration through to safety guardrails and sub-300ms response times at scale.

What you'll be doing:

• Designing and optimizing a multi-phase query processing pipeline with deterministic fast-paths, vector search, and parallel processing
• Integrating and managing Azure OpenAI models for query classification, HyDE rewriting, and safety evaluation
• Managing OpenSearch indices for FAQ, SmPC, and PromoMats content with cosine similarity scoring
• Developing and maintaining a safety guardrails framework covering adverse events, off-label use, PII detection, and prompt injection
• Optimizing LLM prompts, temperature settings, and latency to meet P95 response time targets
• Implementing fallback strategies and circuit-breaker patterns for service resilience
• Designing global resource pooling (boto3 sessions, thread pools, HTTP clients) to minimize per-request overhead

What we're looking for:

• Hands-on experience with Azure OpenAI and LLM orchestration
• Strong Python skills including async and threaded processing
• Practical knowledge of vector search and OpenSearch
• Experience with prompt engineering, evaluation metrics, and performance tuning
• Nice to have: safety guardrails design, embedding model A/B testing, circuit-breaker patterns, FastAPI