AI Engineer (LLM + Voice AI for Enterprise Training & Simulation Platform)

About the project We are building an AI-powered corporate training and skills-assessment platform for large enterprise clients (banking, government, and telecom). The platform replaces static e-learning with interactive, AI-driven practice: employees train against realistic AI roleplay scenarios, get scored automatically against defined rubrics, and receive personalised coaching feedback. We are looking for an experienced AI engineer to own the AI layer of the product end to end. What you'll build AI roleplay simulations — voice and chat based conversational agents that play customers, auditors, or difficult scenarios so employees can practise sales, service, and compliance conversations Automated assessment and scoring - LLM-based evaluation of learner responses against structured rubrics (accuracy, tone, compliance adherence, objection handling), with consistent, explainable scores Personalised feedback and coaching per-attempt strengths/gaps summaries, suggested next modules, and improvement tracking over time Adaptive learning paths - difficulty and content sequencing that adjusts to learner performance Content ingestion pipeline , turning client SOPs, product documents, and policy PDFs into structured training modules, question banks, and knowledge bases (RAG) Analytics , competency scoring, cohort comparison, and manager-facing reporting driven by model outputs Multilingual support , English and Arabic, including correct pronunciation handling in voice modules Required skills 3+ years building production software, with at least 2 years on LLM-based applications (not just notebooks and demos) Strong Python; FastAPI or similar async framework; PostgreSQL Hands-on LLM orchestration: OpenAI / Anthropic / Bedrock / Gemini APIs, tool calling, structured JSON outputs, streaming Serious prompt engineering skill — system prompt design, guardrails, output schema enforcement, handling non-determinism RAG in production: chunking strategies, embeddings, vector databases (Qdrant / pgvector / Pinecone), retrieval quality tuning Evaluation experience: building test sets, scoring LLM outputs, measuring regressions when prompts or models change Real-time voice AI: STT and TTS integration, streaming pipelines, latency optimisation (LiveKit, WebRTC, or similar) Cloud deployment on AWS, Docker, CI/CD, logging and monitoring for AI systems (token cost, latency, failure rates) Clear written English and the ability to work independently with async communication Nice to have Experience with LMS/e-learning platforms, SCORM, xAPI, or assessment engines Self-hosted inference: vLLM, SGLang, Ollama, GPU serving and quantisation Fine-tuning or LoRA training on domain data Arabic NLP or multilingual TTS/STT work Enterprise integrations: SSO/SAML, HRMS, on-premise or private-cloud deployment React/Next.js familiarity to integrate with our frontend team