Vaccine Clinical Trial Literature Synthesizer with CrewAI and Daytona
Public health advisory committees require rapid synthesis of emerging vaccine research and safety profiles across heterogeneous literature. Orchestrate a crew of specialized research agents using CrewAI in isolated Daytona sandbox environments to summarize trials and flag potential study biases. Achieve zero missing key safety signals across benchmark trial corpora.
What you are building
Orchestrate an autonomous multi-agent literature synthesis crew using CrewAI and Daytona isolated execution sandboxes.
How work is evaluated
Evaluates recall of identified adverse event safety signals from vaccine clinical trial papers.
Datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
zero_missed_signals
Ensures all known critical safety signals are extracted.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
safety_signal_recall
Recall metric for detecting documented trial adverse events • target: 1 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Learning goals
Design autonomous agent roles and tasks using CrewAI framework
Deploy code execution tasks safely inside Daytona workspace environments
Extract clinical outcome endpoints and adverse event signals from PubMed articles
Generate unified evidence tables for federal advisory review briefs
Resources
Curated set of 50 PubMed abstracts and clinical trial reports related to childhood vaccine safety monitoring.
Challenge setup
Evaluates recall of identified adverse event safety signals from vaccine clinical trial papers.
Challenge input
JSON containing PubMed text abstracts of 5 vaccine trials.
CrewAI
Framework specified for role-playing multi-agent collaboration.
Daytona
Isolated development environment provider for safe code execution.
Evaluated output
JSON object with safety_signals list, risk_of_bias_assessment, and strategic_summary.
- Ensures all known critical safety signals are extracted.
- Recall metric for detecting documented trial adverse events • target: 1 • range: 0-1
- Benchmark: Clinical Trial Systematic Review Benchmark
- Safety Signal Recall target: 1
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Shows the first task, up to three tools, and up to three evidence items.
View technical recipe
Configured tools
- Daytona · Required
- Daytona · Required
- CrewAI · Optional
- Zed · Optional
Evaluation contract
- zero_missed_signals · Weight 1
- safety_signal_recall · Weight 1
Metrics
- safety_signal_recall · Target 1
Benchmarks
- Clinical Trial Systematic Review Benchmark
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing PubMed text abstracts of 5 vaccine trials.
Output: JSON object with safety_signals list, risk_of_bias_assessment, and strategic_summary.
Execution harness: Python
Evaluation tools
- CrewAI · Multi-agent Orchestration Framework
- Daytona · Agent System Sandbox
Recipe freeze marker
No recipe freeze time is recorded.