Senior LLM / AI Engineer
Ninja
| Company | Ninja |
| Category | Engineering |
| Location | Riyadh |
| Remote | On-site (inferred) |
| Employment | Not stated |
| Level | Senior |
| Salary | Not stated by the employer |
| Posted | 15 Jul 2026 |
| Last verified | 3 Aug 2026 |
| Source | Employer ATS (workable) |
Description
We are looking for AI Engineer to join Ninja and play a foundational role in building our AI function. This position is focused on designing, developing, and deploying production-ready agentic AI systems that deliver real business impact. You will work on customer care automation, knowledge management pipelines, and intelligent targeting and decision-making engines, ensuring scalable, reliable, and high-performance AI solutions. This is a hands-on engineering role for someone who enjoys shipping production systems rather than building experimental prototypes, and who is excited to shape the future of AI at Ninja. Requirements Strong Python (3.11+), with production experience in FastAPI services Agentic workflow design: Has designed or built multi-step agentic workflows, and understands the tradeoffs between deterministic state machines / graphs and open-ended ReAct-style agents. LLM integration: Has built directly against the Anthropic and/or OpenAI SDKs in production. Comfortableworking without heavy abstraction layers LangGraph experience is a plus. Retrieval judgment: Understands when structured/keyed database lookups beat vector search, and hasimplemented RAG when it's actually the right tool Data layer: Comfortable with PostgreSQL, SQLAlchemy 2.0 (async), Alembic migrations, pgvector, and Redisfor session/state management. Production discipline: Has shipped systems with real guardrails: cost controls, latency budgets,escalation/fallback paths, and evaluation loops Familiarity with an LLM observabilitytool (Langfuse, LangSmith, Helicone, or equivalent) is expected. Prompt engineering & evals: Practical experience writing, testing, and iterating on prompts, including building orusing eval frameworks to catch regressions. MCP / tool-serving: Exposure to MCP (Model Context Protocol) servers or similar tool-exposure patterns for AI assistants is a strong plus. Multilingual (nice-to-have): Experience with Arabic or other non-English LLM applications (dialect handling,multilingual evals) Public GitHub required: A public GitHub profile with real, browsable work — personal projects, open-source contributions, or code samples they can point to. Solid hands-on experience: Genuine hands-on builder, not someone who has only architected or managed. They should be able to walk through code they personally wrote line-by-line, including decisions and mistakes. Ownership and proactivity. This is a founding role with minimal oversight. We need someone who identifies problems and drives them to done without being told step-by-step what to do. Riyadh-based (or hybrid). Based in Riyadh, or willing to work hybrid with regular in-person presence in Riyadh.
You found the opening. Now track it.Tracker, radar and AI drafts in one place.erioun.com →