Research Aide– MCS – Ding, Hong – 5.15.26

Argonne National Laboratory•Lemont, IL

13d•Onsite

About The Position

The student will curate a structured domain knowledge corpus — drawing from technical references, historical sensor logs, incident reports, and operational procedures — and generate cross-modal embeddings to populate a vector index that serves as the retrieval substrate for the analyst. They will then build a retrieval-augmented generation (RAG) pipeline that conditions large multimodal models on evidence retrieved from this corpus, designing and evaluating dense, sparse, hybrid, and cross-modal retrieval strategies in which a query in one modality surfaces supporting evidence in another. A major part of their work will be benchmarking the pipeline on anomaly detection, contextual interpretation, and cross-modal reasoning tasks, analyzing how chunking, retriever quality, and retrieved-context configuration affect end-to-end accuracy, faithfulness, hallucination rates, and the marginal value of retrieval over a non-augmented baseline — identifying which retrieval-augmented approaches are most effective for mission-relevant monitoring scenarios.

Requirements

Currently enrolled in undergraduate or graduate studies at an accredited institution.
Graduated from an accredited institution within the past 3 months.
Actively enrolled in a graduate program at an accredited institution.
Must be 18 years or older at the time the appointment begins.
Must possess a cumulative GPA of 3.0 on a 4.0 scale.
Must complete a satisfactory background check.
All students remain subject to applicable drug testing policies.
The entirety of the appointment must be conducted within the United States.

Nice To Haves

Candidates may be required to complete pre-employment drug testing based on appointment length.

Responsibilities

Curate a structured domain knowledge corpus from technical references, historical sensor logs, incident reports, and operational procedures.
Generate cross-modal embeddings to populate a vector index.
Build a retrieval-augmented generation (RAG) pipeline that conditions large multimodal models on evidence retrieved from the corpus.
Design and evaluate dense, sparse, hybrid, and cross-modal retrieval strategies.
Benchmark the pipeline on anomaly detection, contextual interpretation, and cross-modal reasoning tasks.
Analyze how chunking, retriever quality, and retrieved-context configuration affect end-to-end accuracy, faithfulness, hallucination rates, and the marginal value of retrieval over a non-augmented baseline.
Identify effective retrieval-augmented approaches for mission-relevant monitoring scenarios.