Senior ML/NLP Engineer - Upgrade Advanced RAG System for Scientific Document Research

Looking for extensive search/retrieval experience I have a working, well-tested RAG (Retrieval-Augmented Generation) system built in Python for processing 300+ scientific/technical PDF documents. The system already implements state-of-the-art retrieval techniques (hybrid search, hierarchical indexing, reranking, multi-hop reasoning, guardrails, evaluation) and has solid benchmarks. Looking for someone who can take it to the next level — implementing specific upgrades grounded in the latest 2025-2026 research papers to improve retrieval quality and add new capabilities. The work involves: Upgrading existing components with newer algorithms backed by recent peer-reviewed research Adding an agentic retrieval layer with proper reasoning loops Benchmarking alternative approaches for key pipeline stages (parsing, embeddings) on our actual corpus before committing Building a comprehensive evaluation and ablation suite with paper-ready visualizations Writing clean, testable code that follows existing patterns in the codebase This is NOT a build-from-scratch project. looking for- Deep hands-on experience building and optimizing RAG pipelines (not just tutorials — real systems) Familiarity with latest retrieval research (2024-2026 papers) Ability to read a research paper and implement the key ideas cleanly Experience running proper ablation studies and benchmarks Strong Python, PyTorch, Hugging Face Transformers