Adarsh Singh
ML researcher focused on retrieval, tabular QA, and representation learning
Tempe, Arizona
Joined July 2026
Network
1.9K connectionsSummary
Researcher focused on retrieval and tabular question answering: Adarsh develops retrieval systems and representation methods for structured data, exemplified by the CRAFT pipeline and related papers that target efficient, scalable table retrieval and robustness to format variations. adarsh+2
Practitioner who produces open-source code and datasets: actively publishes code, tooling, and LLM-generated enrichment data (CRAFT GitHub and Hugging Face collections) to enable reproducible retrieval experiments and community reuse. github+2
Combines academic research with prior industry engineering experience: transitioned from engineering roles (Oracle, Disney+ Hotstar internship) into graduate research at ASU, bringing practical engineering skills (cloud, encoding/benchmarking, SQL/perf) to research systems development. adarsh+1
Affiliated graduate student at Arizona State University working in CoRAL Lab: holds a formal ASU profile and publishes peer-reviewed and preprint work with collaborators at ASU and partner institutions. asu+1
Work
Education
Projects
Writing
Improving Robustness of Tabular Retrieval via Representational Stability
April 1, 2026Work studying serialization-induced instability in table retrieval and proposing centroid averaging and a lightweight residual bottleneck adapter to produce more serialization-invariant table representations and improve retrieval robustness.
CRAFT: Training-Free Cascaded Retrieval for Tabular QA
May 1, 2025Paper introducing CRAFT, a zero-shot cascaded retrieval approach for open-domain table QA that enriches table representations with LLM-generated titles/descriptions and composes sparse and dense retrieval stages to achieve strong retrieval recall and efficiency.