Virtual Research Collaborations

How to Secure Online Research Experience & Internships to Supercharge Your PhD Applications📚✍️ 1. Explore Research Assistant Gigs in Unconventional Places ✅ Leverage paid research projects – Websites like Kolabtree, Upwork, and Fiverr list freelance research gigs where you can get hands-on experience. ✅ Reverse-search faculty grant recipients – Researchers who just got funding need assistants. Check NSF, ERC, and NIH grant databases for newly awarded grants in your area, then reach out. 2. Tap Into “Ghost” Virtual Labs & Research Networks Most students chase well-known labs. Instead, dig into: ✅ Decentralized research groups – Platforms like The Open Science Network, NeuroMatch Academy, and Complexity Explorer allow you to collaborate on research without formal enrollment. ✅ Virtual research fellowships – Some professors unofficially mentor online researchers. Find labs with past virtual fellows and email PIs about informal mentorship. ✅ Contribute to preprint papers – Read recent arXiv, SSRN, and bioRxiv papers, find research gaps, and email authors with proposed extensions (they often let you collaborate on follow-ups). 3. Turn Your Own Research into a “Published” Internship Struggling to find formal research roles? Create your own. ✅ Conduct a meta-analysis – Find 10-20 papers in your field, extract data, and run a new analysis. Post results as a preprint on OSF, ResearchGate, or arXiv. ✅ Use open datasets – Sites like Kaggle, Zenodo, and Dryad provide real-world data. Publish your own mini-study using R or Python and share it on GitHub. ✅ Collaborate on GitHub research projects – Search “open research” repositories on GitHub and contribute to existing projects. 4. Get Inside Research Groups via Non-Obvious Entry Points Instead of just emailing professors, get on their radar first: ✅ Attend their virtual talks & ask intelligent questions – Many labs post recorded talks on YouTube or university sites. Engage before emailing. ✅ Comment on their latest work on Twitter/LinkedIn – Academics notice engaged readers more than generic emails. ✅ Offer a useful skill – If you know data visualization, programming, or writing, offer to assist in exchange for research exposure. 5. Leverage “Elite” Research Networks ✅ Volunteer for journal editorial boards – Some open-access journals allow grad students to join review panels. Search for “early-career editorial positions” in your field. ✅ Use academic Discord servers – Some of the best research groups have invite-only Discord or Slack communities. Find their members on Twitter and ask how to join. ✅ Apply for online research “sprints” – Some labs hold week-long virtual research projects (e.g., Turing Way Research Sprints). Join and convert participation into long-term work. #PhDApplications #ResearchExperience #OnlineInternships

Death of distance? How Remote Teams Now Drive Disruptive Science In 2023, a landmark study published in Nature by Lin, Frey & Wu suggested that on-site teams are more likely to produce disruptive scientific work. The prevailing rationale was that face-to-face contact and spontaneous interactions are essential catalysts for generating truly novel ideas. However, recent follow-up research by Frey and a co-author published last week in Industrial and Corporate Change reveals a crucial caveat. This dynamic was only true before the widespread adoption of modern digital collaboration tools. As the attached figure illustrates, the trend has reversed. Today, disruptive scientific work is actually more likely to emerge from geographically distributed teams. This shift has profound implications for how we organize R&D teams. In the modern digital age, the evidence suggests that bringing together the best minds, no matter their location, is more important than geographical proximity.

What if our medical AI startup made all of our research public and worked directly with the online Discord community to develop the most useful open models? Kind of scary to have a company's research progress fully transparent, but I think pro-open source, crowd-sourced collective intelligence approaches to research are an under-appreciated resource. We're now giving away compute and focusing on optimizing the Discord "science-in-the-open" workflow. We want to collaborate with volunteers and academics to train great models & publish top papers. Last week we hosted 3 public Google Meets to share our work on fMRI foundation models, real-time brain-to-image decoding, and pathology foundation models. Below are the links to the recordings from these meetings: fmri foundation model: https://lnkd.in/ew8_iWwh pathology foundation model: https://lnkd.in/eWCJBcpA real-time brain-to-image: https://lnkd.in/e5MUXX9g For some more context, I've successfully led a few successful and a few unsuccessful Discord-based research collaborations in the past. I've seen that most of the time these open science projects fail. I'm now also sharing a 10-page blog post on my philosophy behind why Discord collaborations often fall apart and our strategies to ensure that doesn't happen with any of the projects we support at MedARC. https://lnkd.in/eXUdNWU3 We also want to support the medical AI online research ecosystem more generally—if you want to lead your own research project (e.g., as an independent researcher or as an academic in a lab) we are keen to hear from you and we can potentially support you by providing you access to compute, our community, and our structured support to ensure you reach your goals. Join us on Discord: https://lnkd.in/eT7Ed7X5

"Here, we expand the capabilities of LLMs for science by introducing the Virtual Lab, an AI-human research collaboration to perform sophisticated, interdisciplinary science research. The Virtual Lab consists of an LLM principal investigator agent guiding a team of LLM scientist agents through a series of research meetings, with a human researcher providing high-level feedback. We apply the Virtual Lab to design nanobody binders to recent variants of SARS-CoV-2." "The Virtual Lab achieved its goal of engaging in a sophisticated, interdisciplinary science research project, as demonstrated by its design of nanobodies with experimentally validated, diverse binding profiles across multiple strains of SARS-CoV-2. The human researcher and team of LLM agents in the Virtual Lab worked together through a series of meetings to rapidly build a complex nanobody design pipeline that incorporates state-of-the-art machine learning and computational biology tools. Building this pipeline required knowledge of multiple areas of science from immunology to protein folding to machine learning and required making decisions that involved reasoning across many aspects of the project simultaneously. The Virtual Lab successfully built and ran this nanobody design pipeline, starting with a set of four well-characterized nanobodies (Ty1, H11-D4, Nb21, and VHH-72) with potency and diverse binding modes against early variants of SARS-CoV-232–35 and developing them into 92 nanobody candidates for recent variants of SARS-CoV-2 that were experimentally validated by human researchers. These 92 nanobodies—efficiently selected from the trillions of nanobody sequences with one to four mutations—include exciting candidates for further development, such as an Nb21 mutant that enhances binding to the JN.1 [receptor binding domain] RBD and gains binding to the KP.3 RBD and a Ty1 mutant that gains binding to the JN.1 RBD. This outcome serves as an example of how human researchers can partner with LLM agents in the Virtual Lab to rapidly achieve a promising scientific result that can streamline further experiments. Even if the ultimate scientific decisions of the Virtual Lab agents are similar to those in the scientific literature, the ability of the agents to quickly adapt those methods to the scientific question at hand shows how LLM agents can potentially empower human researchers to do complex, interdisciplinary science even when they do not have access to an expert panel of human scientists." https://lnkd.in/ehifq-CF

In a recent Nature paper researchers have developed a virtual laboratory that integrates multiple large language models (LLMs), termed ‘AI scientists’, each assigned specific scientific roles to collaboratively achieve objectives set by human researchers. This system successfully designed 92 nanobodies capable of binding to SARS-CoV-2, with over 90% demonstrating binding affinity to the original virus variant. Notably, two nanobodies also showed potential against newer variants. The virtual lab operates with minimal human intervention, conducting ‘team meetings’ to assess progress and utilizing tools like AlphaFold and Rosetta for protein design. This approach signifies a shift towards human–AI collaboration in interdisciplinary research, highlighting the importance of human oversight to validate AI-generated hypotheses and ensure safety. https://lnkd.in/dWuPduQu