AI for Natural Sciences
Mentor: Sergey Samsonau, PhD
What Students Do
Research in various fields of natural sciences can benefit from automation and novel computational approaches — that's where AI comes in. Students identify real problems, design solutions, build and test them, then document and present findings. No toy datasets. No pre-solved exercises. Real questions that matter.
About the Mentor Dr. Samsonau
- Trained 100+ students in research methodology at NYU
- Built and directed research labs program at PRISMS (one of the top USA high schools)
- Led AI projects at financial, medical, and educational institutions
- Organized AI Meets Science NYC-metro area conference
- Developer of original research education methodologies with focus on applying modern technology to scientific research
Lab Structure
Weekly 1-1.5 hours Zoom meetings with Dr. Samsonau and fellow lab members. Normally 4-6 members per meeting.
Projects: Team-based or individual, depending on scope, interests, and capabilities.
For team projects, members take complementary roles:
- Domain specialist
- AI specialist
- Data analysis specialist
- Engineering specialist
- Coordination and outreach specialist
Three-stage process:
- Formulate — Understand what ML can solve for a specific science problem
- Develop — Build and test solutions
- Deliver — Document and present results
How the Lab Works
Publicly available datasets — Real scientific datasets from researchers, government agencies, and scientific organizations
Self-directed learning — Fundamentals learned independently, then applied together in meetings
Real problems — No toy exercises; questions that matter to actual scientific fields
Project Examples
| Field | Applications |
|---|---|
| Biology | Image classification, genomic pattern analysis, ecological modeling |
| Chemistry | Molecular property prediction, spectroscopy interpretation |
| Physics | Signal detection, trajectory analysis, simulation enhancement |
| Environmental Science | Climate data, sensor networks, species detection |
| Astronomy | Object classification, public telescope archive data |
Prerequisites
- Active SoTS Membership
- Commitment to weekly meetings and independent work
- Willingness to learn tools independently between sessions
- Curiosity about both AI and science
Not for students who need hand-holding or want to watch lectures.
About the Methodology
Developed and used 2021-2024. 100+ students trained, 20+ research collaborations. Now adapted for motivated high school students.
References:
Questions? Contact ains.lab@teenscientists.org