Yifan Xu

Yifan Xu

he/him

Undergraduate student interested in designing optimization algorithms and writing high-performance implementations to solve large-scale linear and nonlinear programming problems.

Enabling Workforce Intelligence through Occupational Taxonomy Alignment

Status: Ongoing (since August 2025) · Advisors: César A. Uribe, Isabella Loaiza-Saa · Affiliations: Rice University & Massachusetts Institute of Technology

Labor-market analysis increasingly relies on synthesizing data across non-comparable occupational taxonomies, and the resulting fragmentation makes it hard to measure equivalence and change, coordinate across global organizations, and compare occupations under technological change. We introduce a network-based occupational translator that represents each occupation as a structured collection of tasks embedded in a global task meta-network, with task relationships estimated via marginal co-occurrence or sparse inverse-covariance (graph-lasso) models.

Occupational dissimilarity is formulated as a Fused Gromov–Wasserstein problem that jointly aligns task semantics (from textual embeddings) and the structure of occupation-specific task networks, yielding soft, interpretable matches. Applied to O*NET U.S. occupations from 2017–2024 and scaled to roughly 900K occupation pairs, the translator recovers about 80% agreement with official crosswalks while exposing task-driven sources of misalignment.

Accepted at the 11th Annual Conference on Network Science and Economics (Miami, April 2026), the 12th International Conference on Computational Social Science / IC2S2 (Burlington VT, July 2026), and the 15th Economic & Financial Networks workshop at NetSci 2026 (Boston, June 2026, oral).

Materials