PharmIQ: platform for pharmacists
AI-powered shift management, inventory control with expiration tracking, and a RAG assistant grounded in MINSAL regulations.
01 Problem and context
A pharmacist manages shifts, inventory with critical expiration dates, and MINSAL regulatory inquiries: each in a separate tool, and regulatory compliance is precisely where an invented answer carries legal consequences.
A pharmacy runs under MINSAL health regulation, and three of its daily tasks carry direct legal weight: expiration control, traceability of who moved which drug, and answering a regulatory question. That last one is where a generic AI tool becomes dangerous: a fabricated answer about a regulation is not a product bug, it is exposure for the pharmacist who trusted it. The rest of the day goes into coordinating shifts under constraints that keep changing.
02 Solution
A single platform: AI-generated shift schedules based on constraints defined by the pharmacist; inventory management with email expiration alerts; and a RAG assistant grounded in MINSAL regulations that cites its sources and explicitly states when something is not covered by the regulations.
- Data model and isolation: Postgres tables with RLS policies, so access is decided at the database and not in the client.
- Inventory: stock, prices, and expiration control by batch.
- Traceability: a record of who entered each drug and when, auditable after the fact.
- Alerts: email notice before a batch expires.
- AI shift scheduling: generated from the constraints the pharmacist defines, regenerated whenever those constraints change.
- Regulatory corpus and RAG assistant: MINSAL regulation indexed; the assistant cites the rule and says when a question falls outside the corpus.
03 How it's built
Pharmacy management platform: AI-generated shift scheduling based on pharmacist-defined constraints; inventory management with expiration tracking and email alerts; and a RAG assistant grounded in MINSAL regulations that provides sourced answers without hallucinating. Real client, in sales conversations; actively used by at least one pharmacist.
04 Ethical decisions
- The assistant answers by citing the MINSAL rule and says plainly when something is not in the regulation, instead of filling the gap with something plausible.
- Data isolation is handled by RLS policies on the Postgres tables: permission is evaluated at the database, where it does not depend on the client behaving well.
- Traceability is by name: every entry is tied to who made it and when. It is auditable, and that includes whoever operates the system.
- AI generates the shifts, but the pharmacist defines the constraints: the decision about people stays with the person accountable for it.
05 Highlights
- RAG assistant grounded in MINSAL regulations: responds with cited sources and zero hallucinations, in a domain where a fabricated answer carries legal consequences.
- AI-generated shift management based on constraints defined by the pharmacist.
- Inventory with expiration tracking and automated email alerts.
- Real client: currently used by at least one pharmacist; in sales conversations.
06 Limitations
The corpus is MINSAL regulation, and that defines how far the assistant reaches: inside the corpus it cites the rule, outside it says it does not have one, and in no case does it replace the pharmacist's judgment or a formal legal opinion. Regulation changes, so the corpus needs periodic re-indexing. Retrieval quality depends on how the regulation is written, and expiration alerts are only as good as the data entered. Today at least one pharmacist uses it: behavior across multiple branches and at higher volume is not proven yet.
The code is private due to its commercial process. I can show the system running in a technical walkthrough: request it by email.