We are seeking an engineer with practical LLM and RAG experience to build a production-ready AI system using Persian-language documents. Project details will be disclosed under a confidentiality agreement.
On-site collaboration in Mashhad is preferred.
Responsibilities
• Build scalable retrieval pipelines using query rewriting, structure-aware chunking, embeddings, hybrid search, reranking, and metadata filters. • Generate verifiable, source-grounded answers with user-selected sources, version-aware retrieval, and appropriate handling of ambiguity, conflicting evidence, sensitive requests, and uncertainty. • Collaborate on document extraction, OCR quality, incremental indexing, failure recovery, and source updates, including invalidation of affected cached and generated content. • Develop multi-step Q&A, long-document summaries, source and case-study comparison, structured extraction, interactive learning, and exercises with explanatory answers and personalized practice. • Manage conversation memory and user context with access, correction, and deletion controls. • Build reproducible evaluations for retrieval, citations, answer faithfulness, completeness, summaries, and exercises; prevent test-data leakage and improve performance through error analysis. • Optimize cost, tokens, latency, and concurrent performance; develop APIs, monitor services, version configurations, and document implementation. • Enforce data isolation and retention policies, control external data sharing, and defend against prompt injection.
Required Skills
• Strong Python, API development, and hands-on RAG experience beyond tutorials. • Practical knowledge of LLMs, context management, structured outputs, tool calling, embeddings, BM25, and reranking. • Experience evaluating models and troubleshooting retrieval and generation errors. • Familiarity with Persian text processing, SQL, document metadata, Git, Docker, logging, and deployment. • Strong problem-solving, technical English, teamwork, and confidentiality practices.
Preferred Qualifications
• Experience with Qdrant, pgvector, Elasticsearch/OpenSearch, LangChain, or LlamaIndex. • Experience with Persian OCR, advanced retrieval, document versioning, learning features, case comparison, or conversational memory. • Experience with fine-tuning, open-source model serving, and production monitoring and load testing.
Deliverables and Collaboration
Deliver a maintainable, documented, API-accessible RAG core with verifiable responses, controlled knowledge updates, and measurable quality, cost, and performance. Work with data, backend, and content specialists; this role owns the AI components and their integration.
Application
Please include your location, availability for on-site work in Mashhad, a relevant project and your contribution, and a brief example of your evaluation approach or a technical challenge solved. High-level descriptions are sufficient for confidential projects.
گروه حقوق آسیا، ارائه دهنده آموزش های حضوری و آنلاین در قالب صوتی، تصویری و جزوات برای دانشپذیران آزمونهای حقوقی (وکالت، اختبار، قضاوت، سردفتری، ارشد و دکتری) است.