Overview
Senior AI Engineer (m / w / d) - RAG & Search Optimization - Freelance
Responsibilities
- Optimize the Retriever: Dive deep into our multi-step retriever, fine-tuning the interplay between Reciprocal Rank Fusion (RRF) for similarity search and BM25 for keyword relevance in OpenSearch.
- Build an Evaluation Framework: Design and implement a robust testing suite for our conversational search from the ground up. Your framework will be the source of truth for measuring precision, recall, and other key relevance metrics.
- Develop Test Data: Curate and create high-quality, domain-specific test datasets to rigorously benchmark and validate retriever performance.
- Enhance the Ingestion Pipeline: Analyze and improve our complex document ingestion process to ensure data is indexed effectively for optimal retrieval.
- Collaborate and Innovate: Work in a dynamic environment, responding to ad-hoc requests and proactively identifying opportunities to improve our system, potentially by introducing advanced techniques like rerankers.
Qualifications
- Expert-Level JavaScript / TypeScript: Proven experience building and maintaining production systems with NestJS.
- RAG & Search Experience: Demonstrable, hands-on experience building or optimizing RAG pipelines. You must be comfortable with the theory and practice of hybrid search.
- OpenSearch Proficiency: Strong practical knowledge of OpenSearch (or Elasticsearch), including query DSL, indexing strategies, and performance tuning for search relevance.
- AI / ML Fundamentals: A solid grasp of core AI concepts, particularly in NLP, vector embeddings, and search evaluation metrics (precision, recall, etc.).
- Pragmatic & Proactive Mindset: You thrive in a fast-paced environment, are comfortable with ambiguity, and can take initiative to build solutions from scratch.
- Nice to have:
- Hands-on experience with AI orchestration frameworks like LangGraph.
- Familiarity with advanced retrieval techniques, such as implementing rerankers (e.g. cross-encoders).
- Direct experience with BM25 and fusion methods like RRF.
Details
- Location: remote
- Start: 06.10.2025
- Duration: till 31.12.2025
- Workload: Full-Time (200 Hours)
- Language: English