← Cases
BrazilHealthcare and life sciences2026 – ongoing

Biofy

Genome-based infection diagnosis with AI

Biofy cut time to a treatment recommendation from five days to four hours

WorkingVendor's case studyVendor ↗

Problem

Standard PCR plus antibiogram testing is slow and fails often, which delays the right antibiotic.

Approach

Biofy runs full-genome DNA sequencing and AI models on Oracle Cloud with 32 NVIDIA H100 GPUs to identify bacteria and predict antibiotic resistance. Its Abby engine suggests treatments for clinical workflows.

Results

5 days to 4 hoursTime to a treatment recommendationVendor's case study[1]
95-97%Diagnostic accuracy, compared with 80% for traditional testingVendor's case study[1]

Takeaway

Biofy used GPU computing to shorten a multi-day lab step to hours. The accuracy figures come from the company and vendor.

Categories

Evidence
Vendor's case study

Sources

  1. Biofy: from testing to treatment
    NVIDIA2026Vendor publication

Source summaries

[1]NVIDIA2026

NVIDIA describes Biofy's sequencing and AI pipeline on cloud GPUs. It reports that time to a treatment recommendation fell from five days to four hours, with diagnostic accuracy of 95–97%.

Full details are in the original sources.

Added 25 Oct 2026 · Updated 25 Oct 2026
Open case page ↗