Devanshu Singh

Education

Johns Hopkins University — MSE, Data Science (05/2022), GPA: 3.6
Concentration in Language and Speech
Johns Hopkins University — BS, Applied Mathematics and Statistics (05/2020), GPA: 3.4
Double Major: BA, Interdisciplinary Studies

Relevant Coursework: Machine Learning, Data Mining, Introduction to Statistics, Natural Language Processing, Vision as Bayesian Inference, Deep Learning, Introduction to Nonlinear Optimization

Certifications: UCLA HIPAA Certification; All of Us Data Certification
University of Washington — PhD Candidate, Political Science (09/2024 – Present)

PhD Candidate
University of Washington | Seattle, WA | Sep 2024 – Present

Led discussion sections (~30 students) for political science courses with intensive writing components.
Researched emerging technology and AI governance; co-authored political science article published in Isonomia Quarterly.

Data Analyst
Center for Advanced Defense Studies (C4ADS) | Washington, DC | Jan 2023 – Jul 2024

Built ETL pipelines to scrape and ingest text data, collecting 10+ TB of trade, corporate, shipping, and other records for PAI/OSINT investigations.
Created pipelines for entity extraction, resolution, and data modeling of PEPs, shell companies, sanctions lists, and other compliance-related entities.
Owned data governance standards: documented lineage/metadata, integrated multiple data streams, and standardized data models.
Developed RAG prototypes for unstructured corporate records (LangChain, LlamaIndex, OpenAI).

Product Manager
Mindgram.ai | Washington, DC | May 2021 – Aug 2021

Built a custom annotated corpus of medical documents with Prodigy (1k+ labeled examples of disease and clinical entities) for NER and relation extraction.
Fine-tuned a pre-trained transformer model to ~90% accuracy converting unstructured clinical docs into structured data.
Directed an engineering intern and led deployment of the NLP pipeline into production to enable scalable text mining and market access insights.

Biostatistics Volunteer
UCLA | Los Angeles, CA | Jul 2025 – Present

Performed multivariate regression analysis with Python and Stata on CDC WONDER mortality data.
Produced 2 tables of annual predicted trends for lung cancer mortality for a co-authored manuscript in preparation.

Research Assistant
Johns Hopkins University Dept. of Political Science | Jan – Dec 2022 | Baltimore, MD

Co-authored papers and chapters on the impact of AI/internet on democracy, presented at ISA-NE.
Raised $2,000 in research funding from interested parties.

Research Assistant
Neurosymbolic Computation Lab, Johns Hopkins University | Baltimore, MD | May 2021 – May 2022

Designed and trained custom Transformer architecture with PyTorch for computational linguistics research.
Scaled model size and data to 10+ GB using parallel processing on GPU clusters.

Technical: Python (Pandas, NumPy, PySpark, PyTorch, LangChain, LlamaIndex), R (tidyverse), SQL, AWS, Databricks, Spark, Docker, EHR management (NextGen)
Governance & Compliance: Data lineage, bias & fairness evaluation, HIPAA compliance, audit-ready pipelines
Domain Knowledge: Healthcare analytics, OSINT, political science applications of AI
Languages: English (Native), Hindi (Native), Spanish (Beginner)

Perez N, Singh DR, Singh RR. “Lung Cancer Mortality, 1999–2024: Disproportionate Increase in Premature Mortality in Certain Subpopulations in the 2020s.” Manuscript in preparation.
Yen EY, Singh DR, Singh RR. “Trends in Systemic Sclerosis Mortality Over Forty-Eight Years, 1968–2015: A US Population-Based Study.” Arthritis Care & Research 73 (2021): 1502–1510. doi:10.1002/acr.24411.
Singh RR, Singh DR, Yen EY. “Worsening premature death burden gap from systemic sclerosis in men and black persons: A US nationwide population-based study.” J Scleroderma Relat Disord. 2023;8(1):20–26. doi:10.1177/23971983221140538. PMCID: PMC9896199.
Deudney D, Singh D. “Bounding Power: The Artificial Superintelligence Control Problem, Public Safety, and Republican Constitutionalism.” Isonomia Quarterly. Link.