Vacancy: Senior AI Engineer (LLM / Product AI)

Company: DevCasi Location: Cyprus Employment: full-time, remote format Salary Range: 3000-4000 USDT Industry: IGaming

About the Role We are a B2B platform for iGaming, where our clients launch their own products. The AI direction covers internal tools and the client-facing part: working with analytics, product management via API, recommendations, virtual characters, and other user scenarios. The role involves high technical autonomy and participation in architectural decisions. Task details and priorities will be discussed during the interview.

Responsibilities

• Develop LLM agents for working with data from ClickHouse and PostgreSQL.
• Connect agents to the platform API so they can help configure the product and perform actions.
• Design secure execution of actions: permission checks, parameter validation, confirmation of critical operations, auditing, and rollback.
• Ensure strict isolation of data, context, and available actions between tenants.
• Develop RAG over product documentation.
• Work on client-side AI features: recommendations, hints, virtual characters, and conversational scenarios.
• Build evals and regression sets to control the quality of agent responses and actions.
• Make architectural decisions: what to use off-the-shelf, what to develop independently, and where self-hosted inference is needed.
• Integrate AI services into the existing infrastructure: Kubernetes, Kafka, CI/CD, and observability.

Requirements

• 5+ years of development experience and practical experience launching LLM systems into production.
• Python and experience operating services under real load.
• Practical experience with RAG, tool calling, and agents performing actions via API.
• Understanding of LLM system limitations and typical failure scenarios.
• Experience building evaluation pipelines.
• SQL and experience with analytical DBMS: ClickHouse, BigQuery, Snowflake, or similar.
• Understanding of LLM application security: prompt injection, access control, and preventing cross-tenant leaks.
• Understanding of principles for building reliable agent systems: authorization, action confirmation, idempotency, auditing, and handling partial failures.
• Independence — tasks come as "it needs to work," not as a ready-made technical specification.
• English B2+.

Will be a plus: experience in regulated domains — iGaming or fintech; recsys/ranking; Go at code reading level; Kubernetes and Kafka; understanding of GPU inference economics.

Tools We expect experience with some of this stack:

• Agents: LangGraph, OpenAI Agents SDK, DSPy, Pydantic AI.
• Evals and tracing: Langfuse, LangSmith, Braintrust, DeepEval, Arize Phoenix.
• Gateway and routing: LiteLLM, Portkey.
• RAG: pgvector, Qdrant, Weaviate, rerankers.
• Inference: vLLM, SGLang.
• Integrations: MCP. The specific set is not important — it's important to understand why each layer is needed and what will happen if it's removed.

Our Stack Python, Go, Kubernetes, Argo CD, Kafka, ClickHouse, PostgreSQL, Grafana/Loki, GitLab CI.

Conditions Real data and load, mature engineering infrastructure, budget for models and experiments.

Application Send your resume and 5–10 sentences about one AI system you brought to production: what it did, how quality was measured, what broke after launch, and how you fixed it. Contact: @Nbagama