Next Epoch 2024

Last week, NextEpoch 2024 brought together a diverse group of college students to explore the intersection of machine learning and biological research. Held from August 5–7, 2024, the workshop covered a range of topics and included coding sessions that demonstrated the essential role of AI in RNA structure prediction.

Here at Vast.ai, we're proud to be a NextEpoch sponsor, providing not only the prizes this year but also the GPUs that powered the participants' work – whether they were coding online or on-site at Rouskin Lab.

This year's program, led by Harvard Medical School Assistant Professor Silvi Rouskin and co-sponsored by the Burroughs Wellcome Fund, was a resounding success!

Perhaps most importantly, the three-day workshop furthered NextEpoch's goal of advancing science by bringing machine learning to the next generation of researchers.

Inside the NextEpoch Experience

Designed particularly to engage first-generation college students and underrepresented minorities, NextEpoch aims to show how machine learning (ML) can be applied beyond traditional tech roles. With a focus on active learning, the program empowers participants to explore new applications for ML in research settings.

To that end, NextEpoch 2024 was packed with insightful discussions, hands-on exercises, and a healthy dose of competition in a hackathon at the end. Here's how it unfolded:

Day 1

The program started off with a lecture on the fundamentals of machine learning, followed by a practical exercise in writing gradient descent algorithms. The highlight of the day was a guest lecture from Google DeepMind's John Jumper, the first author behind AlphaFold2. His talk on “Structure Prediction of General Biomolecular Interactions with AlphaFold” left a lasting impression.

Day 2

The second day focused on diving deeper into the history and architectures of machine learning. After an introduction to PyTorch, students built their first models using the open-source machine learning library. Yves Martin from Atmo also delivered an interesting talk on ML applications in weather prediction, adding another layer of real-world relevance.

Day 3

The day of the Kaggle competition! Given that understanding RNA structure is key to understanding its function, participants were tasked with building a model to predict RNA secondary structure based on its nucleotide sequence.

After some intense competition and outstanding efforts from all participants, winners were crowned. The top three teams won GPU credits in varying amounts as a prize from Vast.ai.

Celebration and Reflection

A huge congratulations to all of the winners this year!

• First place went to Nav Mon with an F1 Score of 0.63519.
• Second place went to Team SJD (Sheriff Wahab, Jimena Luque, and Daniel Ukaegbu) with a score of 0.61813.
• Third place went to The Maple Roos' Nathaniel Dalton with a score of 0.60544.

An honorable mention also went to The Wobble Bases (with an F1 Score of 0.59248), recognizing the strong performance of team members Anastacia Parks, Thomas Costa, and Isabella Nieblas.

For our part, we'd like to thank all those who contributed to making NextEpoch 2024 an unforgettable experience.

As we look back on this year's workshop, we're reminded of the incredible potential that AI and machine learning hold for the future of science. At Vast.ai, we’re committed to supporting initiatives like NextEpoch that are driving the future of AI-powered scientific advancements.

Additionally, by providing access to high-performance GPUs at an affordable price through our cloud GPU rental marketplace, we aim to empower researchers, students, and developers alike to tackle complex problems and accelerate their work. Whether you're competing in a hackathon, conducting research, or training AI models, Vast.ai is here to provide the compute power you need.

Here’s to many more epochs of discovery and innovation in the years to come!