RIFT 2026  ·  HealthTech Track  ·  Explainable AI

Precision Medicine
Powered by Pharmacogenomics

PharmaGuard analyzes your patient's genetic variants to predict drug metabolism risks — delivering clinically actionable recommendations before a single prescription is written.

100k+
Preventable deaths/yr
6
Critical genes
6
Supported drugs
CPIC
Guideline aligned

Adverse drug reactions are
silently killing patients

Most of these deaths are preventable. The missing piece is knowing how a patient's unique genome affects how they metabolize drugs — before they take them.

100,000+
Americans die from ADRs every year

Adverse drug reactions are one of the leading causes of preventable death in the United States. The majority occur in patients with known pharmacogenomic risk factors that were never tested for — because the tools to do so at scale simply didn't exist at the point of care.

🧬
Genetic Variants Change Everything

A single nucleotide polymorphism in CYP2D6 can turn a standard dose of Codeine into a lethal overdose for ultra-rapid metabolizers, or render it completely ineffective for poor metabolizers. Genetics aren't optional context — they are the answer.

Clinicians Need Answers Now

Doctors don't have time to decode raw VCF files or cross-reference CPIC tables. PharmaGuard translates raw genomic data into clear, structured, plain-language clinical recommendations in seconds.

From genome to
clinical decision

A four-step pipeline that transforms raw VCF files into CPIC-aligned dosing recommendations with LLM-generated clinical rationale.

01
Upload Patient VCF File
Drag-and-drop or select a standard .vcf (v4.2) file up to 5MB. PharmaGuard validates format, extracts INFO tags (GENE, STAR, RS), and confirms file integrity before processing.
VCF v4.2 · Up to 5MB
02
Pharmacogenomic Variant Identification
Our parser scans for clinically significant variants across 6 critical genes — CYP2D6, CYP2C19, CYP2C9, SLCO1B1, TPMT, and DPYD — and maps star allele diplotypes to phenotype classifications (PM, IM, NM, RM, URM).
CYP2D6 CYP2C19 CYP2C9 SLCO1B1 TPMT DPYD
03
Drug Risk Prediction
Enter a drug name (or multiple, comma-separated). PharmaGuard cross-references the patient's diplotype against CPIC guidelines to output a structured risk label with confidence score and severity.
Safe Adjust Dosage Toxic Ineffective Unknown
04
LLM-Generated Clinical Explanation
Claude synthesizes the variant data, phenotype, and risk label into a structured clinical summary — citing specific rsIDs, explaining biological mechanisms, and providing dosing rationale aligned with CPIC guidelines. Output is downloadable as structured JSON.
CPIC Aligned · Structured JSON · Copy to Clipboard

Everything a clinician
needs to act

Designed for precision, speed, and clarity — every feature exists to translate genomic complexity into confident clinical decisions.

VCF File Parsing
Robust parser handles standard VCF v4.2 with INFO tag extraction. Validates format and provides user-friendly error messages for malformed files.
Risk Classification
Five-tier risk labeling system (Safe → Toxic) with confidence scoring and severity levels, aligned with CPIC evidence grades.
LLM Explanations
AI-generated clinical summaries cite specific variants, explain metabolic pathways, and provide actionable dosing language a doctor can act on.
Structured JSON Output
Every result exports a schema-compliant JSON object — downloadable and clipboard-ready for integration into hospital EHR workflows.
Multi-Drug Support
Analyze multiple drugs simultaneously with comma-separated input. Each drug generates an independent risk profile from the same genomic data.
CPIC Alignment
Every recommendation cross-referenced against Clinical Pharmacogenomics Implementation Consortium guidelines — the gold standard for pharmacogenomic decision support.

Built on a serious stack

Production-grade tools chosen for reliability, speed, and clinical-grade data handling.

Backend
Flask
Python · REST API
AI / LLM
Claude API
Anthropic · Sonnet
Database
PostgreSQL
SQLAlchemy ORM
Genomics
PyVCF
VCF v4.2 Parsing
Auth
Flask-Login
Werkzeug · Hashed
Frontend
Jinja2 + CSS
Glassmorphism UI
Guidelines
CPIC
v1.x · Evidence A/B
Deploy
Render
Public · Live URL

Built by people who
care about the problem

A multidisciplinary team bringing together genomics, clinical knowledge, and engineering.

A
Satyam Maddeshiya
Full-Stack Engineer
B
Sanskar Gupta
ML / LLM Engineer
C
Raj Kamal Jha
Genomics & Bioinformatics
D
Aryan Saini
Motivator

Ready to make prescriptions
safer?

Upload a VCF file and get a full pharmacogenomic risk report in seconds — no genomics expertise required.