All case studies
Pharma/RegTech

Vero

Catches 40%+ of citation issues before a human reviews it
2025
TEAM
4 engineers, 1 domain analyst
DURATION
12 months
TECH
React, Node.js, Python (NLP pipeline), PostgreSQL, PageIndex RAG, AWS

The Challenge

Pharmaceutical companies face a grueling compliance bottleneck known as MLR (Medical, Legal, Regulatory review). Every marketing material (webinar slides, product brochures, social posts) must be reviewed by medical, legal, and regulatory teams before it can go live. For one client, a single product brochure was taking 14–21 days to clear MLR.

The problem wasn't the reviewers. It was the process. Materials were passed around via email chains. Version control was non-existent. Comments were scattered across PDF annotations, Slack threads, and in-person meetings. There was no audit trail, no parallel review, and no way to track whether a material was compliant, rejected, or pending.

Our Approach

We built a compliance workflow platform that treats MLR as a structured process rather than an unstructured approval chain. The system ingests marketing materials (PDFs, slide decks, video scripts), auto-extracts claims and assertions, then routes them through configurable review workflows.

The key innovation was our reference-checking pipeline. Marketing claims in pharma literature must cite supporting references (clinical studies, FDA labels). Our system automatically parses claims from the material, cross-references them against a connected knowledge base of approved references, and flags any claim that lacks a valid citation, before a human reviewer even sees it. This uses our PageIndex approach for fast, auditable lookups without hallucination risks.

For reviewers, we built a unified review interface that consolidates comments, version diffs, and compliance status in one place. Parallel review became the default: medical, legal, and regulatory reviewers working simultaneously, with the system tracking conflicts and escalations automatically.

How the review pipeline actually works

Every piece of marketing material moves through the same automated pipeline before a human ever opens it: the pipeline described in the sections above, laid out end to end:

Ingest

PDFs, slide decks, and video scripts uploaded into the platform.

Extract claims

NLP pipeline parses every marketing claim and assertion from the material.

Cross-reference

PageIndex RAG checks each claim against the approved reference knowledge base.

Auto-flag

Claims missing a valid citation are flagged, before a human reviewer sees the file.

Parallel human review
Medical

Clinical accuracy and safety review.

Legal

Contractual and liability review.

Regulatory

FDA / compliance review.

Approved + audited

Every decision logged with a full audit trail: who reviewed what, when.

Outcomes

  • MLR review cycle reduced from 14–21 days to 3–5 days for standard materials
  • Pre-review claim validation catches 40%+ of citation issues before human review
  • Full audit trail for every material: who reviewed what, when, and their decision
  • Parallel review workflows eliminated email-chain bottlenecks
  • Compliant with pharma regulations (FDA 21 CFR Part 11, HIPAA)