TraceRx and Immunity

Antibiotic Resistance Prediction Tool3
Antibiotic Resistance Prediction Tool 2
Antibiotic Resistance Prediction Tool
Main site pic 2
Main Site pic 1

Inspiration In hospitals, patients move a lot ER, floor, ICU, sometimes another hospital. Important lab results and past infections often get lost in paperwork or never get checked against the medicine that was ordered. We wanted a tool that helps teams keep that story straight, and warn when antibiotic resistance might still matter.

What it does TraceRx helps staff go from a lab report to a clear review: capture the report, check it, match it to the order, and save what was decided. It can catch problems like “the lab is final, the medicine is still ordered, but nobody linked a review.” Immunity sits inside that flow. It looks at past colonization, old culture results, and recent antibiotics, then explains in plain steps why resistance risk may be higher right now.

How we built it: We use a vision model to read messy lab images and PDFs, then simple rules to double-check the answers so mistakes don’t slip through. Python/Streamlit runs the main workflow; React helps with the review screens. Immunity adds the resistance reasoning on top. We only demo with fake/synthetic patient data.

Challenges we ran into: AI can misread a blurry report, so we couldn’t trust it alone we needed rules and human review. There’s no single national patient ID in the U.S., so sharing a patient’s full history across hospitals is hard. Building the lab workflow, Immunity logic, and UI at the same time under hackathon pressure was tough.

Accomplishments that we're proud of: We connected the full path: messy report → checked fields → review → documented decision. Immunity doesn’t just show a number; it shows why. We kept the product honest: it supports doctors and stewards it doesn’t replace them.

What we learned Missing paperwork can hurt patients as much as a wrong clinical guess. Best approach: AI reads, rules verify, people decide. Old infections and antibiotics matter less as time passes, so “when” matters. Knowing who the patient is and knowing whether a record changed are two different problems.

What's next for TraceRx and Immunity: Connect more tightly to real hospital systems (orders and lab results), with easy human override. Work with pharmacists to tune Immunity’s timing and drug coverage. Longer term: help hospitals keep continuity across sites while patient charts stay inside the hospital.