MedAssist AI-Powered Medication Dispensing Agent

Inspiration

Medication errors remain one of the most preventable yet persistent safety issues in healthcare. Nurses and pharmacists often work in high-pressure environments where a simple mistake, such as selecting the wrong drug or dose, can have serious consequences. Our inspiration was to design a system that removes human error from the most critical step: the physical act of dispensing medication. We wanted to create an intelligent robotic assistant that can see, reason, and act safely before a drug reaches the patient.

What it does

MedAssist is an AI-powered robotic medication verification and dispensing system. It scans a medication tray using a camera, identifies each vial with computer vision, and verifies the medication against patient records. The system performs strict safety checks such as drug identity verification, dosage matching, allergy conflict detection, and high-alert medication handling. If everything matches, the robot dispenses the medication. If any mismatch or risk is detected, the system immediately stops and alerts medical staff.

How we built it

We built MedAssist using a layered sense → reason → act architecture. A vision model processes the camera feed and detects medication vials. The detection results are passed into a deterministic rule engine that validates the medication against patient orders stored in a database. A Toolhouse agent generates operational summaries and logs system events. Supabase is used as the database to store patient data, scan results, decisions, and system logs. A Smallest.ai voice agent communicates decisions to staff in real time. Finally, a Cyberwave robotic arm executes the dispensing action only after all safety checks pass.

Challenges we ran into

Integration between Supabase and Toolhouse Couldn't test it on live robots Somewhat time went into learning new softwares

Accomplishments that we're proud of

We successfully built a full end-to-end prototype that demonstrates how AI agents, computer vision, and robotics can work together in a safety-critical environment. The system can detect medication trays, validate drugs against patient orders, explain its decisions, log events in a database, and control a robotic arm for dispensing. Most importantly, MedAssist demonstrates how technology can proactively prevent medication errors before they reach patients.

What we learned

Through this project we learned how to design AI systems for safety-critical workflows. We gained experience integrating computer vision APIs, AI agent platforms, real-time voice interfaces, and robotics. We also learned the importance of building deterministic rule layers alongside AI models to ensure reliability and explainability in healthcare applications.

What's next for MedAssist AI-Powered Medication Dispensing Agent

Next, we want to expand MedAssist into a fully deployable clinical assistant. Future work includes integrating hospital electronic health records (EHR) systems, improving vision accuracy with larger medication datasets, building a real-time monitoring dashboard for hospital staff, and expanding the robotic system to handle larger medication inventories. Ultimately, we envision MedAssist as a reliable AI assistant that helps hospitals reduce medication errors and improve patient safety at scale.