Staff AI Engineer – Agentic LLM Systems

About the role

Versant, which owns StockStory (now part of CNBC), is building AI-powered equity research for individual investors. It is looking for a Staff AI Engineer to turn LLM capabilities beyond prototypes into robust, production-grade systems. The role suits engineers who enjoy designing end-to-end workflows, setting technical direction and building reusable foundations, and offers direct access to experienced market professionals.

The work covers orchestration, retrieval, tool use, evaluation and operational safeguards, and the engineer helps define engineering standards and architectural patterns. The position is on-site in Prague, Czech Republic, and the candidate is employed through an Employer of Record; day-to-day work follows Versant's structure.

Key responsibilities

• Design, build and ship LLM-powered systems integrated into new and existing workflows, covering orchestration, retrieval, tool use, evaluation and operational safeguards.
• Build end-to-end distributed agentic AI solutions that are reliable and scalable.
• Lead cross-functional initiatives requiring coordination across multiple teams and shared systems.
• Mentor other engineers and define quality standards and best practices.
• Own the long-term maintainability, observability and cost efficiency of AI systems.
• Monitor model quality drift, prompt alignment, alerting and operational health.

Required profile

• Bachelor's degree in Computer Science or equivalent practical experience.
• At least 5 years of experience designing, building and maintaining software systems.
• Genuine interest in investing, public markets and fundamental business analysis.
• Prior experience leading an AI systems team.
• Experience shipping and operating multiple GenAI or LLM systems in production.

Required skills

• Strong backend expertise in at least one strongly typed language, preferably TypeScript.
• Solid understanding of cloud computing primitives, especially AWS.
• Understanding of agentic AI concepts such as tool use, function calling, state machines or graphs, retrieval and reranking, structured outputs, memory, guardrails and evaluation.
• Experience developing and deploying agentic workflows with frameworks such as LangChain, Mastra or LangGraph.
• Hands-on experience debugging scaled LLM systems and incident response.
• Experience building human-in-the-loop evaluation pipelines and evaluation suites for agentic systems.
• Experience with RAG, embeddings and model fine-tuning.