Dr. Sarah P.
Research Scientist · PhD Stanford · ML research automation
Spent 7 years doing ML research — a PhD at Stanford on large-scale optimization, then 3 years publishing at NeurIPS and ICML. Frustrated by how slow the research cycle is, I started building agents to automate the parts that don't require intuition: literature review, experiment logging, paper writing, and ablation analysis. These agents are what I wish I'd had during my PhD.
Experience
Research Scientist · Google DeepMind
2022 – 2024Large-scale optimization and training efficiency. Co-authored 6 papers published at NeurIPS and ICML.
Postdoctoral Researcher · Stanford AI Lab
2020 – 2022Scalable training methods for foundation models.
Education
- 2020
Stanford University
PhD, Computer Science — Machine Learning
- 2015
Seoul National University
BS, Computer Science
Awards & Recognition
- 2023
Best Paper Award
NeurIPS 2023
- 2016
NSF Graduate Research Fellowship
National Science Foundation
Agents by Dr. Sarah P.
Research Paper Writer
Give it your experimental results and get back a conference-format paper draft, covering experiment design, ablations, related work, and references. ## What it does - **Conference-format drafts:** results shaped into a paper in the expected structure - **Experiment & ablation sections:** methodology and ablations written from your data - **Related work:** prior art positioned around your contribution - **References:** citations formatted to the venue ## Where it fits - **A grad student writing up:** "Draft the methods and ablations from these results." Sections ready to edit. - **A researcher prepping a submission:** "Structure this into a conference paper." A formatted draft to refine. - **A team racing a deadline:** "Write the related-work section around these citations." A positioned draft to tighten. ## How it works 1. **Give it your results:** data, settings, and the contribution. 2. **It drafts the paper:** structure, ablations, related work, and references. 3. **You get a draft back:** conference-format sections to edit and submit. Built for graduate students and researchers. Every draft is yours to verify, edit, and stand behind.
Academic Writing Assistant
Give it your research direction and existing materials and get back first drafts of each paper section, formatted with citations. ## What it does - **Section drafts:** intro, methods, results, and discussion drafted from your materials - **Citation formatting:** references handled in the expected style - **Academic structure:** the conventional shape of a paper applied - **Material-grounded:** built from the notes and sources you provide ## Where it fits - **A grad student facing a blank page:** "Draft the introduction from my notes." A section to edit instead of start. - **A researcher organizing:** "Turn these results into a methods and results section." Structured drafts to refine. - **A writer formatting:** "Format these references and weave them in." Clean citations in place. ## How it works 1. **Give it your direction and materials:** the topic plus notes and sources. 2. **It drafts section by section:** structured academic prose with citations. 3. **You get drafts back:** sections to verify, edit, and assemble. Built for graduate students and researchers. Every draft is yours to fact-check, edit, and stand behind.
Technical Concept Explainer
Give it a system, a code diff, or a data table, and get back a shareable HTML page that explains the concept visually. ## What it does - **Visual explainers:** a concept turned into an annotated HTML page with diagrams and callouts - **Diff visualization:** a code change rendered as a clear before/after a non-author can follow - **Data storytelling:** a table or dataset turned into charts and a guided narrative - **Shareable output:** self-contained HTML, client- and doc-ready ## Where it fits - **An engineer briefing stakeholders:** "Explain how our caching layer works for a non-technical audience." A visual page anyone can follow. - **A reviewer walking through a change:** "Visualize this refactor diff for the team." A clear before/after to share in the PR. - **An analyst presenting findings:** "Turn this results table into an explainer." Charts plus narrative on one page. ## How it works 1. **Hand it the material:** a system, a diff, or a data table. 2. **It builds the visual:** diagrams, annotations, or charts with a guided explanation. 3. **You get an HTML page back:** share it as is or refine the layout. Slash commands: `/generate-visual-plan` `/generate-web-diagram` `/generate-slides` `/fact-check` Built for technical docs and client briefings. Every page is a draft you can edit and reshare.
Literature Review & Paper Writer
Give it a research direction and get back a literature review, paper draft, and formatted references, covering the full path from survey to submission. ## What it does - **Literature review:** the field surveyed and summarized - **Paper drafting:** sections written toward a submission - **Reference formatting:** citations checked and formatted to the venue - **End-to-end:** the path from survey to submission in one workflow ## Where it fits - **A researcher starting out:** "Survey the literature on this topic." A review to ground the project. - **A student writing up:** "Draft the paper and format the references." Sections plus clean citations. - **A team near a deadline:** "Check our citations and tighten the draft." A submission-ready pass. ## How it works 1. **Give it a direction:** the research topic. 2. **It runs the workflow:** review, drafting, and references. 3. **You get outputs back:** a literature review and a formatted draft to refine. Slash commands: `/ars-full` `/ars-citation-check` `/ars-abstract` `/ars-format-convert` Built for the full research-to-submission workflow. Every draft is yours to verify and stand behind.