Autonomous Scientific Discovery Agent
Inspired by groundbreaking collaborations in AI-driven drug discovery, this challenge tasks you with building an autonomous scientific research agent. Your agentic system will simulate the initial phases of drug or gene therapy development by autonomously reviewing scientific literature, generating novel hypotheses, and outlining experimental designs. Emphasize the use of a multi-agent framework to enable specialized roles (e.g., 'Literature Reviewer,' 'Hypothesis Generator,' 'Experimental Designer') that collaborate to achieve complex scientific goals. The system should be capable of processing vast amounts of information and presenting actionable insights.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a multi-agent system using CrewAI and Mixtral 8x22B for autonomous scientific literature review and hypothesis generation.
How work is evaluated
The evaluation will assess the agent system's ability to autonomously generate coherent scientific hypotheses and design experiments based on provided research questions. Emphasis will be on the quality of reasoning, novelty of hypotheses, and clarity of experimental steps.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Master CrewAI for orchestrating role-based, goal-oriented multi-agent workflows for complex tasks.
Implement advanced prompting techniques with Mixtral 8x22B to extract, synthesize, and generate scientific hypotheses from textual data.
Design and manage experiment configurations and parameter tuning using Hydra for reproducible scientific AI research.
Deploy and serve specialized AI models (e.g., protein folding, molecular dynamics simulators) via RunPod for on-demand scientific computation.
Orchestrate complex data ingestion, transformation, and analysis pipelines using Prefect to support agent decision-making and scientific output generation.
Reference links and supporting material
Sample data for 2 tasks
Challenge setup
The evaluation will assess the agent system's ability to autonomously generate coherent scientific hypotheses and design experiments based on provided research questions. Emphasis will be on the quality of reasoning,...
Challenge input
{'research_question': 'string', 'literature_keywords': ['string']}
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{'hypotheses': [{'title': 'string', 'statement': 'string', 'justification': 'string'}]}
- Checks if the output is a valid JSON object matching the expected schema.
- Ensures at least 2 distinct hypotheses are generated.
- HypothesisNoveltyScore target: 0.7
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- Checks if the output is a valid JSON object matching the expected schema.
- Ensures at least 2 distinct hypotheses are generated.
Metrics
- HypothesisNoveltyScore · Target 0.7
- ExperimentalDesignFeasibility · Target 0.8
- RelevanceToResearchQuestion · Target 0.85
Workflow tasks
- Task 1
Input: {'research_question': 'string', 'literature_keywords': ['string']}
Output: {'hypotheses': [{'title': 'string', 'statement': 'string', 'justification': 'string'}]}
Execution harness: Python
- Task 2
Input: {'hypothesis': 'string', 'goal': 'string'}
Output: {'experiment_title': 'string', 'objective': 'string', 'steps': ['string'], 'expected_outcomes': ['string']}
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.