A small PhD reading group at Emory focused on natural language processing, large language models, information retrieval, and ML applied to health and clinical settings. We meet every other week to present papers, build quick prototypes, and explore new tools together.
Sessions rotate between paper presentations, vibe coding challenges, and group lab explorations. Topics are shaped by the research interests of everyone in the group, with a particular focus on clinical NLP, LLM evaluation, RAG, vision-language models, and alignment.
Interested in joining? Reach out via Slack.
| Date | Type | Session |
|---|---|---|
| Week of May 18 |
Paper |
Kickoff — intro presentation
Organizers present. First 10 min covers ground rules, format overview, and interest survey results. Paper is something broadly accessible, e.g. a foundational RAG or LLM survey. Presenter slots for the summer assigned here.
|
| Week of Jun 1 |
Paper |
Rotating presenter #1
Paper nominated and voted on by the group via Slack. Suggested topic pool: LLM evaluation, health NLP, VLMs in radiology. Paper shared at least 5 days before the session.
|
| Week of Jun 15 |
Vibe code |
Vibe coding session #1
Prompt released ~2 weeks prior. Pairs welcome. Each person or pair demos in 3–4 minutes. Closes with a 10-min group debrief on approaches.
|
| Week of Jun 29 |
Paper |
Rotating presenter #2
Suggested topic pool: LLM alignment, thinking traces, evaluating models without them knowing they're being tested, RAG improvements.
|
| Week of Jul 13 |
Group lab |
Group lab session
Theme chosen from the interest survey. Options: integrating deep research tools into a publishing workflow, working with new HuggingFace benchmarks, or building a custom annotation tool. Each person shares one finding or demo (5 min), then open discussion.
|
| Week of Jul 27 |
Paper |
Rotating presenter #3 & retrospective
Final session of the summer. Last 10 minutes: quick retrospective on format, topics for fall, and whether to continue. Good moment to collect paper nominations for a potential fall season.
|
One presenter walks the group through a paper: motivation, method, key results, and their own take. Group discussion follows.
A research problem prompt goes out ~2 weeks early. Everyone builds a small solution (solo or in pairs) and demos it to the group.
Everyone explores the same theme independently (a tool, dataset, or workflow) and brings one finding or demo to share.