Inspiration
“Behind every clinical trial is a patient waiting.”
Every year, thousands of promising drugs are delayed, not by science, but by paperwork. The manual, error-prone process of Clinical Trial Data Review is a critical bottleneck that stalls medical progress. We saw that a single day of delay can cost over $8 million, but more importantly, it's a day a patient waits for a life-saving therapy.
Our inspiration became intensely personal when we modeled the journey of a synthetic patient, Oliver (SONA301-MRC-002), a 43-year-old in a trial for a new Type 2 Diabetes drug. Her progress was stalled by a data backlog where a late-signed consent form and out-of-range lab values went undetected for days, putting both her safety and the trial's integrity at risk. This led us to ask: why should a paperwork error delay a patient's journey? Why must patient safety wait for manual checks?
HealFast AI is our answer.
What it does
HealFast AI is an AI-powered agentic framework designed to automate the validation of critical clinical documents, ensuring data is accurate, compliant, and ready for submission in near real-time. It transforms the slow, linear data review process into a proactive, intelligent, and parallel workflow.
Our system:
- Ingests a chaotic mix of unstructured PDFs (Lab Reports, ICFs, AE forms)
- Extracts and Structures the data into a unified, machine-readable format.
- Validates every data point against a complex, study-specific protocol rulebook for accuracy and completeness.
- Verifies the data for adherence to critical regulatory and ethical standards like ICH Good Clinical Practice (GCP).
- Escalates any discrepancies or complex edge cases to human experts via an Action Center for a final decision.
- Generates clean, analysis-ready datasets and human-readable compliance summaries.
How we built it
We designed HealFast AI on the principle of specialized agentic intelligence, where dedicated AI agents handle focused tasks for maximum accuracy and speed.
Validation Agent (Protocol Expert):
Extracts structured data from unstructured clinical documents and applies complex, protocol-driven rules to validate accuracy, completeness, and consistency.Compliance Agent (GCP Guardian):
Performs a secondary review to ensure adherence to regulatory and ethical standards, like correct consent timing and Good Clinical Practice (GCP) compliance.Orchestration & Human-in-the-Loop:
Using n8n, we built a workflow engine to coordinate agents, trigger validations, and escalate only edge cases to human experts — ensuring high accuracy without bottlenecks.
🛠️ Technology Stack
- Platform: n8n
- Techniques & Approaches: Regex Parsing, Prompt Analysis, JavaScript
Integrations: Gmail, Dropbox, Google Sheets -**LLM : Gemini
Input: Real-world simulated clinical documents
Output: Validated JSON + Record Status
Challenges we ran into
Data Chaos: Normalizing dozens of variable PDF formats.
Engineering Trust: Making AI explainable for regulated GxP environments.
Orchestration: Designing seamless collaboration between specialized agents.
Accomplishments that we're proud of
Our proof-of-concept for HealFast AI demonstrated transformative results that directly address the industry's biggest pain points:
85%+ faster patient-batch review — cut review time from 40-60 hrs → under 6 hrs.
Achieved 98%+ first-pass accuracy, reducing rework significantly.
Escalated Serious Adverse Events 90% faster — detection went from 24+ hrs → <1 hr.
Delivered a true multi-agent system solving a mission-critical, real-world bottleneck.
What we learned
- Operational inertia, not a lack of science, is often the biggest bottleneck in drug development.
- Every data point represents a patient. A single delayed signature can have a real impact on a person's trial journey.
- True agentic automation isn't just about speed; it's about creating trustworthy, compliant, and intelligent systems that empower human experts. The goal is to elevate human oversight from tedious checking to strategic decision-making.
What's next for HealFast AI - Accelerating Therapies, Saving Lives
Our vision for HealFast AI extends beyond this hackathon. Our roadmap is focused on turning this powerful proof-of-concept into a truly transformative platform.
Short-Term Goals:
- Implement a Feedback Learning Loop: Allow agents to learn from human corrections in our Action Center to improve the accuracy of extraction and validation models over time.
- Expand Rulebook & Document Support: Increase the library of validation/compliance rules and train DU models on a wider variety of clinical documents.
- Plug into Sponsor Workflows: Develop connectors for major EDC systems and eTMF platforms.
Long-Term Vision (Vision 2026 - From Labs to Lives):
- Global Adoption to Reduce Trial Timelines: Our ultimate goal is to contribute to a 30% reduction in overall clinical trial approval timelines by eliminating the data review bottleneck.
- Enable Predictive Data Integrity: Evolve the agents from reactive error-finding to proactively identifying sites or processes at high risk of producing data errors.
- Accelerate Approvals: By providing regulators with "certified," trustworthy data, we aim to build a faster, more reliable bridge from scientific discovery to the patients who are waiting.
Built With
- dropbox
- gemini
- gmail
- google-sheets
- javascript
- n8n
- prompt
- regex

Log in or sign up for Devpost to join the conversation.