AI Engineer (LLM Workflows & LLMOps) - Muhimma Digital

Full Description

Own the build-out and day-to-day optimisation of agentic LLM workflows that ingest large volumes of qualitative text (survey verbatims, social posts, transcripts) and deliver production-grade data and insights for dashboards. You will translate research requirements into modular chains/agents, optimize and maintain existing workflows, and monitor performance in line with emerging LLMOps best practices

Key Responsibilities:

• LLM Integration & Prompt Engineering

• Design, test, and maintain workflows against OpenAI, Google, or open-weights models via SDKs or REST.

• Help build multi-agent graphs (e.g., LangGraph) that route tasks such as classification, summarisation, and extraction

• NLP Pipeline Development

• Fine-tune or embed open-source models (e.g., sentence-transformers) to tag sentiment and themes in Arabic/English text at scale

• Optimise vector-DB retrieval (Pinecone/Chroma) for hybrid RAG scenarios used by research analysts.

• Cross-Functional Enablement

• Pair with data engineers to keep ETL/ELT pipelines LLM-ready; write minimal documentation and how-to notebooks for researchers.

• Act as a bridge between the research and data teams to understand key requirements and translate them into scalable solutions

Required Skills & Experience:

• 1–3 years hands-on Python (FastAPI, LangChain/LlamaIndex)

• Familiarity with cloud GPU instances (AWS Bedrock/SageMaker, Azure OpenAI, or Runpod)

• Basic DevOps: Docker, git,

• Applied NLP: tokenisation, embeddings, text classification; exposure to Arabic language models is a plus.

• Comfortable reading research papers/blogs and translating them into PoCs quickly.

• Strong written & spoken English; Arabic proficiency preferred.

Nice-to-Have:

• Experience with Streamlit or Dash for quick internal prototypes.

Prior work analysing survey verbatims or focus-group transcripts.