Research

The research case for a PhD in artificial intelligence: directions I want to work on, and the infrastructure, papers, and computational studies behind them. Each entry states its type and how AI was used.

Research Directions

I am applying to PhD programs in artificial intelligence. The work on this site is the case for it: research infrastructure at 12.2M rows, working papers, open machine-learning tools benchmarked against published baselines, and production AI systems that keep a human in the loop. Quantum computing and biophysics show up here as application areas where the machine-learning question is sharp; the doctoral work I want sits in AI itself.

R1 Machine learning for scientific discovery

Models that make a scientific measurement cheaper or a hypothesis testable: protein stability prediction, Markov state models of molecular dynamics, ADMET property prediction, and quality control for cryo-EM and electrophysiology. Benchmarked against published baselines, CPU-only, open source.

R2 AI systems for research knowledge

Retrieval and generation grounded in primary sources: hybrid BM25 and dense retrieval over full-text papers, answers that carry their citations, and a 12.2M-row provenance-tracked graph of the research-funding economy underneath.

R3 Evaluation and human-AI workflows

How to measure the output of AI-assisted research: verification levels, disclosure of the human and AI division of work, agent orchestration with review gates, and what degrades once a person stops checking.

R4 Learning methods for quantum computing

Neural decoders for quantum error correction inside a real-time budget, learned search over quantum-optimization circuits, and how to partition a workload across CPU, GPU, and QPU.

These are active study questions rather than completed results. Each one links to code, a benchmark, or a working paper where one exists.

Portfolio

I use AI as a research and engineering instrument. I define the questions, scope, and evaluation criteria; orchestrate the workflows; review intermediate and final outputs; and make the final technical and editorial decisions. Each artifact states its verification level and the division of human and AI work.

Research AtlasReproducible research infrastructure
Can the global research-funding economy, grants, funders, organizations, people, and the outputs they produce, be reconstructed as one normalized, provenance-tracked graph, open enough for anyone to audit or extend?
~1.67M grants · ~192.7K organizations · ~1.44M people · 75 funders · ~$1.04T normalized funding · ~12.2M graph rows
The structure of public research fundingPreprint / working paper
How concentrated is public research funding across institutions, what fraction of funded work is co-funded across multiple funders, and how does funding translate into measurable output across agencies?
887,016 grants · Gini coefficient 0.929 institutional concentration · ~25% of funded works co-funded across multiple funders
The transformer paper-recommendation advantagePreprint / working paper
Does a transformer-based recommender (SPECTER) beat a classical TF-IDF baseline at recommending related papers, and does that advantage hold across the whole literature or only at the top?
2026
What each funder fundsPreprint / working paper
How specialized is each funder's field portfolio, how much do funders complement rather than duplicate each other, and how stable are those portfolios over time?
2026
Public funding and researcher careersPreprint / working paper
Is the productivity gap between funded and unfunded researchers a causal effect of funding, or mostly selection of already-productive researchers into funding?
2026
The Knowledge-Access GradientPreprint / working paper
Every prior technology widened access to consuming knowledge, to read, learn, and reach what others discovered. None widened access to producing it. Can AI become the first scalable channel that helps people cross that boundary?
78,326 observations · 219 countries · 30 indicators
Neural Decoders for Quantum Error Correction
Can learned decoders match minimum-weight perfect matching (MWPM) on quantum error correction codes while staying inside a real-time decoding budget?
2026
Quantum Similarity and Kernel Estimation
Can swap-test and Hadamard-test overlap estimation reproduce classical cosine similarity and kernel computations on real quantum hardware, and where does the classical cost of state preparation eliminate any advantage?
2026
Academic Paper Ranking and Recommendation SystemAcademic project
Rank academic papers by citation-graph influence and combine that ranking with content-based recommendation.
2022
US States Obesity AnalysisAcademic project
How do state-level obesity rates relate to poverty, income, and other socioeconomic variables, and how strong is that relationship?
2023