AI Engineer

Overview

Key Responsibilities

• Design, deploy, and optimize LLM-powered applications in production environments
• Build and maintain Retrieval-Augmented Generation (RAG) pipelines with robust retrieval and ranking strategies
• Implement scalable inference systems using high-throughput serving frameworks
• Develop and fine-tune transformer-based models using parameter-efficient techniques
• Ensure high-quality outputs through evaluation frameworks and custom benchmarking
• Handle multilingual data processing, with a focus on Arabic and RTL text handling
• Collaborate with cross-functional teams to integrate AI solutions into business applications
• Implement guardrails for safety, compliance, and output control

Core Technical Skills (Must-Have)

• Proven experience deploying open-source LLMs (e.g., Llama, Falcon, Jais, Mistral) on GPU-based infrastructure
• Strong hands-on with inference frameworks such as vLLM, Text Generation Inference, or similar
• Production experience with RAG pipelines: chunking, embeddings, retrieval, reranking, and query optimization
• Expertise in vector databases (pgvector, ChromaDB preferred; also Weaviate, Qdrant, Milvus)
• Experience with multilingual embeddings and RTL language processing
• Fine-tuning expertise using LoRA, QLoRA, or other PEFT methods
• Experience with LLM evaluation tools (RAGAS, TruLens) and building custom evaluation workflows
• Strong programming skills in Python with PyTorch and Hugging Face ecosystem
• Experience with frameworks like LangChain or LlamaIndex
• Deep understanding of NVIDIA GPU environments (CUDA, multi-GPU inference, parallelism, NVLink)

Preferred Qualifications

• Experience building or working with MCP (Model Context Protocol) servers
• Hands-on exposure to regional LLMs such as Falcon or Jais
• Strong background in Arabic NLP (normalization, tokenization, dialect handling, diacritics)
• Advanced prompt engineering (few-shot, structured prompting, instruction tuning)
• Experience implementing AI safety/guardrails (e.g., NeMo Guardrails, Guardrails AI)
• Kubernetes experience managing GPU workloads and NVIDIA operators
• Experience with Document AI tools (e.g., PDF rendering, Puppeteer) supporting RTL/LTR formats