virtualryder/hcls-ai-agents

HCLS AI Agent Suite

Governed AI Agents for Life Sciences — Built on AWS

The agents are not the product. The governed platform that makes them deployable, auditable, and compliant is.

A large systems integrator deploying AI in a pharmaceutical, biotech, medtech, or CRO environment cannot hand a customer a collection of LLM calls and call it done. Every regulated artifact an agent touches — an ICSR report, a submission section, a CAPA record — carries data-integrity, e-signature, and accountability obligations that exist before the first line of agent code is written. This suite embeds those controls from the first commit: deny-by-default authorization, PHI masking, grounding verification, prompt version pinning, a human gate that is framework-enforced (not merely documented), and a tamper-evident audit trail that satisfies 21 CFR Part 11.

The result is a deployable accelerator — not a certified product — that gives an SI engagement team a credible, compliant starting point across eight high-value life-sciences workflows.

Repository status (current): all 8 agents built to flagship depth · 8 AWS-native rebuilds (Strands + Step Functions) · a live Amazon Bedrock + real-connector reference path (Agent 02) · 452 automated tests passing with no API key · CloudFormation quick-deploy + Terraform parity · executive deck, 5-slide customer teaser, and one-page leave-behind included.

---

Positioning

| What this is | What this is not | |---|---| | A governed, auditable accelerator — bring your own LLM call without the compliance scaffolding and you still have a prototype | A certified, validated, production-ready SaaS product you can hand to a customer unchanged | | Eight agents with shared platform controls that compound across the portfolio | Eight point tools built independently with no governance consistency | | A reference for Amazon Bedrock AgentCore Gateway + Identity + Runtime semantics — testable locally, deployable on AWS | A vendor lock-in