ClearPath

Canva presentation

View our Canva presentation

Inspiration

Most of our team is international, and between us we've watched the same story play out in different countries: a patient waiting on a medicine that's stuck somewhere. It might be held at a border over a labeling rule, stalled in customs, or caught behind a wildfire or flood. The medicine exists, but it can't reach the person who needs it.

That gap is bigger than it looks. When patients can't take their medication as prescribed, whether because of shipping delays, cost or access, public health research calls it medication non-adherence. According to data cited by Truveta, it's linked to more than 125,000 deaths a year in the U.S. alone. In clinical trials the stakes are even higher, because a missed dose can hurt the patient and the trial they're part of.

We can't stop a wildfire, rewrite a country's regulations or speed up customs. So we took the tools in front of us this weekend, including satellite data, AI agents and real-time messaging, and built ClearPath.

What it does

ClearPath keeps clinical-trial drug shipments on track from the warehouse to the patient:

• Plans the schedule: it calculates projected delivery, safe ship-by dates, slack and risk from transit times, customs timing and holidays.
• Checks the paperwork: a regulatory agent pulls cited requirements from source text, compares labeling rules across countries and flags when regulations change.
• Watches the route from space: every route is checked against NASA FIRMS fire detections from the MODIS and VIIRS satellites, delivered through NASA LANCE near-real-time data, and against GDACS global disaster alerts. Routes are drawn on Google Maps.
• Finds another way: when something threatens a shipment, the planner calculates alternative plans. A person always approves the change, and nothing changes on its own.
• Keeps the team in the loop: Max, our assistant, works in the web app and over iMessage through Photon. It sends alerts and handles approvals with role-based checks.
• Tracks regulations in every country: ClearPath keeps a saved copy of each country's labeling, storage and import rules, cites where every requirement came from, and flags the affected shipments when a rule changes.
• Keeps every record in one place: Inventory, shipments, schedules, plans, approvals and messages all live in one database, with who did what and when, so every reroute or delay can be traced.

How we built it

• Website: a plain HTML, CSS and JavaScript frontend with a Node server that runs the map, Google Routes and the NASA fire monitor.
• Planner API: a Python service for planning and decisions, with Chalk feature definitions for delivery dates, slack and risk. Results are checked against local formulas, with a recorded fallback when Chalk isn't available.
• Messaging and decisions: the Max agent, with Claude powering cited regulatory answers and Photon handling iMessage.
• One source of truth: a single SQLite database holds proposals, the current plan, approvals, evidence, conversations and the full audit history.
• Grok: we used Grok while building the global shipment map and visualization, and Grok Imagine to create our demo video.
• Calculates risk with Chalk: Chalk feature definitions work out each shipment's projected delivery, safe ship-by date, slack and risk from transit times, customs timing and holidays. ClearPath queries the deployed Chalk features, checks the answers against its own formulas, and records when it falls back to local calculations.

At its core, every shipment comes down to one question: how much room do we have before a patient misses a dose?

$$\text{slack} = \text{deadline} - (\text{ship date} + \text{transit} + \text{customs})$$

When slack runs low or a hazard shows up on the route, ClearPath steps in.

Challenges we ran into

• Oceans don't have roads: Google Routes only returns driving routes, so overseas shipments failed at first. We built an estimated air or sea corridor that's still checked against live disaster data.
• Quotas at 3 AM: we used up Google's daily route limit mid-build, so we added caching and graceful fallbacks to keep the app working.
• Being honest with data: a tool that patients' care depends on can't pretend. We only show a live position when real GPS data exists, label everything else as estimated, and never let an alert change a plan without human approval.
• Putting it together: joining the website, planner, Chalk features and messaging agent across branches meant a lot of merge conflicts and late-night debugging.

What we learned

• How to turn NASA's near-real-time satellite data into practical warnings for logistics.
• How to design human-in-the-loop AI: the system suggests, people decide, and every decision is recorded.
• How to combine Chalk, Photon, Claude and Grok into one product so that no single service breaking takes the whole thing down.

What's next

We've seen this problem up close, and we want ClearPath to grow beyond a hackathon project: live GPS from carriers, more disaster and weather sources, deeper label checks for each country, and a pilot with a real clinical-trial supply team. The disruptions won't stop, but patients shouldn't have to pay for them. And, possibly implement it for usage by pharmaceutical companies for shipments in the countries that have been affected by this problem since a long time.

Built With

• and-grok-and-grok-imagine-for-building-the-app-and-making-the-demo-video-**messaging:**-photon-(imessage)-**maps-and-data:**-google-maps-javascript-api
• chalk
• claude
• css
• gdacs
• google-maps
• google-routes-api
• grok
• html
• javascript
• map-and-fire-monitor
• nasa-firms
• nasa-lance-and-gdacs-**testing:**-python-`unittest`
• node.js
• photon
• playwright-with-chromium
• pytest
• python
• sql
• sqlite