We’re halfway through the Bio x AI Hackathon, and $25K in midpoint prizes have been awarded to three teams for their progress on Agentic Science prototypes.
Over the past few weeks, builders have been prototyping agentic systems, automating wet lab workflows, and designing the backbone of decentralized biotech.
The results are outstanding. And there’s still more time to submit your proposal. Final submissions officially close June 8, and there’s a remaining $100K in prizes for the winners!
Here are the three teams awarded midpoint prizes for showing what's possible when scientific research meets agentic automation.
$10K Prize: Rare Compute’s PDB-MCP Server & METEOR – The Agentic BioML Stack
Rare Compute is pushing the boundaries of agentic biomedicine with two complementary projects: the PDB-MCP Server and METEOR, their experimental BioML system.
PDB-MCP Server: Structured Protein Data for Autonomous Agents
The PDB-MCP Server (Protein-Data-Bank Model Context Provider) is an open micro-service that lets AI agents tap the RCSB PDB (the world’s largest 3D protein archive) through a standards-based MCP interface. Instead of relying on cumbersome downloads or screen scraping, BioML agents can now access high-quality protein structure data through a lightweight FastAPI microservice. Plus, it’s fully compatible with Anthropic’s Model Context Protocol (MCP) spec.
Designed for decentralized deployment via Docker or CLI, the PDB-MCP Server empowers researchers, educators, and startups to host their own mirrors without central gatekeepers.
The project tackles a critical bottleneck in the biotech research process: BioML (Biological Machine Learning) models need curated 3D data. Traditional downloads are bulky and rate-limited, which slows down design loops in fields like immunotherapy, oncology, and enzyme engineering.
The PDB-MCP server tackles this challenge with compact context bundles that include title, method, resolution, and ligands, so reinforcement learning design agents can iterate in seconds instead of hours.
Newsletter
Get the latest updates from the Bio ecosystem
New launches, research progress, and ways to get involved across the Biosphere.
The team built the entire system based on four core principles:
Open: MIT-licensed code, CC-BY-4.0 data.
Inclusive: Runs locally on a laptop or in Kubernetes.
Composable: Plugs into any MCP-aware LLM.
Reproducible: Built-in provenance for every query.
This democratizes access to one of the most critical knowledge bases in biology, leveling the playing field for global BioML research and opening the door to frictionless scientific collaboration across geographies and institutions.
METEOR: Integrating ESM3 for Precision Medicine
Building on that foundation, Rare Compute is integrating ESM3 into METEOR, their system for simulating enzyme variants directly from patient genomes.
Developed by EvolutionaryScale, ESM3 is one of the most advanced protein language models ever released. It’s trained with over 10²⁴ FLOPs and 98 billion parameters, and can unify sequence, structure, and function into a single generative system.
Unlike earlier tools that produce static predictions, ESM3 generates 3D molecular structures and forecasts how mutations impact folding and function. For rare metabolic diseases, where a single point mutation can shift the course of treatment, this level of precision is transformative.
Together, PDB-MCP and METEOR represent a powerful shift in biotech R&D. They give both humans and AI agents frictionless access to structured protein data and the tools to reason over it. This means faster hypothesis generation, better functional predictions, and a stronger foundation for applications like de novo drug design, variant interpretation, and personalized medicine.
With this project, Rare Compute is enabling a new class of open, decentralized, AI-native research workflows that could reshape how biotech gets done.
$10K Prize: SpineDAO’s Chronos – Surfacing Clinical Memory from the Past
Imagine if we had a time machine for biomedical knowledge. We could recover centuries of lost information to understand modern medicine.
Built by SpineDAO, Chronos is a modular, agentic system that uncovers overlooked clinical insights buried in historical spine surgery manuscripts and traditional Indian medicine, especially Siddha and Ayurveda. These systems represent generations of observational data, therapeutic approaches, and anatomical insight, much of it written in classical Tamil, and largely inaccessible to modern science. Most of it isn’t even in digital format!
Chronos changes that.
The team’s prototype digitizes scanned manuscripts, performs OCR, runs named entity recognition using OpenAI and Camel, and constructs a real-time semantic knowledge graph in Neo4j. From there, agents reason over the graph to generate hypotheses and plug into Eliza OS for autonomous downstream execution.
Chronos vision goes beyond simple information retrieval: → Connect the forgotten edge of historical medicine with the frontier of autonomous research. → Recover non-Western medical traditions and give them a first-class seat in AI-native biotech. → Build a long-context agent system that doesn’t just “know things” but helps the scientific system remember.
Chronos could radically broaden the biomedical search space. In spinal care alone, it opens the door to rediscovering treatment strategies, diagnostics, and hypotheses that have been filtered out of the mainstream evidence base, not because they don’t work, but because they were never digitized or translated.
Taking it one step further, Chronos offers a new pattern for the broader biotech industry in general: → Time-aware agent pipelines that treat historical, multilingual, and unstructured knowledge as assets, not liabilities. → Traceable semantic infrastructure for connecting traditional knowledge to modern trials, protocols, and publications. → Decentralized knowledge stewardship via OriginTrail, ensuring openness, provenance, and reproducibility across time.
$5K Prize: ValleyDAO’s Phlo: AI-Powered Platform that Translates Research to Profitable Businesses
ValleyDAO is building Phlo, a platform that helps scientists turn research into real-world biotech companies, powered by AI.
The problem: researchers generate high-potential discoveries every day, but most never leave the lab. The path from experiment to company is unclear, slow, and filled with friction, especially for scientists without access to startup expertise, investors, or translational infrastructure.
Phlo is designed to fix that. It combines scientific analysis with commercial reasoning, giving researchers an AI partner that understands the technical details of their project and helps shape it into a viable, fundable venture.
At the center of the platform is the concept of Artificial Research Intelligence (ARI). ARI goes beyond chatbots or pitch-deck generators. It’s a long-context agent that parses publications, grant proposals, protocols, and founder input to assess feasibility, surface risks, suggest business models, and guide project evolution over time.
The team’s goal is to build a system that integrates:
Scientific document parsing (from preprints to patents)
DAO-native coordination tools (for funding, governance, and protocol execution)
Phlo helps scientific builders focus on what they do best (discovery) while supporting them through the messy middle between idea and execution. For ValleyDAO and other BioDAOs, it unlocks a more efficient, inclusive pipeline for funding early-stage biotech.
This is the kind of infrastructure that can lower the barrier to company creation, make grant-to-startup transitions more fluid, and help the next generation of biotech companies emerge from open science.
We have seen so many great projects emerge from the hackathon, and the best part is: We’re not even done yet! Final projects are due June 8, and $100,000 in prizes are still up for grabs.
If you’ve got a prototype in progress, now’s the time to test, tune, and ship. And if you're just joining? There’s still room. You don’t need permission — you just need a working build.
