When was the last time you wondered if taking Tylenol and Advil together was safe? Preventable adverse drug events (ADEs) cause devastating patient harm every year. Polypharmacy, the act of taking multiple medications at once- is incredibly common, yet patients are left entirely alone to manage complex schedules and possibly dangers drug-to-drug interactions. re-medy is a scalable support engine which doesn't just track pills; it actively intercepts duplicate ingredients, warns the user of drug conflicts, and also produces a schedule for taking medications.
Users bypass tedious manual entry, by snapping or uploading a photo of their medication, the system parses the data finding the exact dosage, active ingredients, and instructions.
Before a medication is saved to the user's calendar, the backend cross-references its pharmacological profile against all other saved medications. Creates a confidence score regarding the usage of each drug combination, if the confidence score exceeds a certain threshold, warnings may take place.
If there is a minor-moderate drug confidence score, it doesn't just throw a warning. The Agentic organizer automatically computes a safe ScheduleSuggestion, dynamically shifting the dosage times to safely seperate the medication. Major Conflicts will also result in a warning, but the medication may still be added if the user insists on it.
When a user feels unwell after administering a drug, they can make log a new symptom. The system will synthesize their active schedule, recent ledger history, and known-side effect rules to rank the most staistically likely "culpript". Further actions also may be suggested.
re-medy is powered by Gemini's 2.5 Flash. It operates as a core Python microsesrvice that outputs JSON contracts to prevent medical hallucinations and unecessary data.
- Vision-to-JSON Pipeline: We use Gemini's multimodal image analysis to scan physical pill bottles, drug images and labels. It instantly extracts messy, real-world data directly into our backend's rigied
ParsedMedicationCondidateJSON schema. - Grounded Triage: Instead of letting the AI guess, we feed Gemini our
side_effect_rulesdatabase alongside the user's active medical ledger. This forces the 2.5 Flash model to rank side-effect culprits based strictly on real data.
❤️ Thank you BACSA hacks for hosting this amazing Hackathon, can't wait to run it back next year.
+++++++++++++
+++++++++++++++
+++++++++++++++++ %##%%%%%%%%
###+++++ +++++####%%%%%%%%%%%
####+++++ ++++####% %%%%%%% +++
#####+++++ +++++#% %%%%% +++
##### ++++++++++++++++# %%%%% +++
#### ++++++++++++++++ %%%% + ++++ +++ + ++ +++ ++++ ++= +++ ++ +
#### ++++++++++++++ %%%% +++++++++ +++++++++++ ++++++++++ ++++++++ ++++++++++ ++++++++++++++++ +++
##### +++++ +++++ %%%%% +++ ++++ ++++ ++++ ++++ ++++ ++++ +++ +++ +++++ +++ +++
#####+++++ ++++++ %%%%% +++ +++ +++ ++++++++++ +++ +++ +++ +++ +++ +++ ++++ ++++ +++
#####*+++ ++++++ %%%%%% +++ ++++++++++++ +++++++++ +++ +++ +++ ++++++++++++ +++ ++++ +++++++
######+ ++++++* %%%%%% +++ +++ +++ +++ +++ ++++ +++= ++++ +++++
####### ++++++ %%%%%%% +++ +++++++++++ +++ +++ +++ +++++++++++ +++++++++++++ ++++
###%%% *+++++#%%%%%% +++ ++++++++ +++ +++ +++ ++++++++ +++++++ +++ ++++
%%%%%% +++%%%%%% +++
%%%%%% #%%%%% +++
%%%%%% %%%%%% ++
%%%%%%%%%%%%
%%%%%%%%%
%%%%%%
%%