Adobe
Machine Learning Engineer (Contract)
Feb 2026 - May 2026 · San Jose, CA
- Built Python graph-RAG research system processing 500+ PDFs with hierarchical ingestion and LLM-based triple extraction.
- Implemented graph traversal retrieval across 20K+ entities to support multi-hop reasoning over domain-specific documents.
- Designed retrieval evaluation harness against baseline RAG, reducing token usage by 32% and improving accuracy by 18%.
Research contract on retrieval-augmented generation for long-form internal docs. Built a graph-based RAG system that hierarchically ingested 500+ PDFs, extracted typed triples with an LLM, and traversed the resulting entity graph at query time to answer multi-hop questions traditional flat RAG missed.
pythongraph-ragllmevaluation