AI / LLM Application Engineer
About the role
We're looking for an engineer to add governed AI features to an existing internal client-relationship/business-development platform, AI-assisted summaries, next-best-action suggestions, and drafted communications, all grounded in approved data, cited, and reviewed by a human before anything is sent. This is a controlled, compliance-conscious build for a regulated business environment, not an open-ended AI experiment.
What you'll build
- RAG-based retrieval over approved internal content, with source citations, freshness flags, and missing-data warnings
- AI-generated account summaries, meeting-prep briefs, and draft follow-ups (with mandatory human approval before anything goes out)
- Versioning, audit logging, and an evaluation/testing process to catch quality regressions
- Support for Arabic-language generation and translation
- Deployment on private/client-controlled infrastructure (not standard cloud AI APIs)
Must-haves
- 4+ years of production software engineering (Python and/or TypeScript)
- 18+ months building AI/LLM features actually used by real users (not just demos or side projects)
- Hands-on experience building RAG systems: ingestion, chunking, embeddings, retrieval, and citation of sources
- Experience designing ways to test/evaluate AI output quality without a large labeled dataset
- Experience with authentication, access control, logging, and human-approval workflows for sensitive data
- Comfortable deploying AI systems on private/self-hosted infrastructure (not just calling a public AI API)
Nice-to-haves
- Experience with Arabic-language AI or right-to-left text handling
- Background in banking, wealth/asset management, or other regulated industries
What's out of scope
- No autonomous sending or external client-facing AI
- No use of public cloud AI services (e.g., Azure OpenAI) , client data stays on approved infrastructure
- No AI model training/fine-tuning from scratch
If you have what it takes to build this solution and are interested, hit that APPLY button!