francosimon

Not hiring, but actively helping healthcare AI startups navigate compliance challenges.

If anyone here is working on health AI and running into HIPAA compliance walls (PHI handling, audit requirements, BAAs with vendors), happy to share what I've learned from consulting with 12+ startups in this space.

The compliance tax on healthcare AI is brutal, but there are patterns and tools that make it manageable.

Feel free to reach out if you're in this domain - always interesting to compare notes on real-world compliance implementations.

VLayer (my project) scans healthcare codebases for HIPAA compliance issues before they reach production. One thing I learned building it: developers rarely think about encryption until it's too late. Tools like VeraCrypt solve the "data at rest" problem, but the bigger issue in healthcare software is unencrypted data in logs and API responses — stuff that's much harder to audit manually.

VLayer (my project) scans healthcare codebases for HIPAA compliance issues before they reach production. One thing I learned building it: developers rarely think about encryption until it's too late. Tools like VeraCrypt solve the "data at rest" problem, but the bigger issue in healthcare software is unencrypted data in logs and API responses — stuff that's much harder to audit manually.

Hey HN – I built NEXUS because discovering A2A agents is only half the problem. The other half is: once you find an agent, how do you trust it? How do you pay it? How do you coordinate multi-step workflows across frameworks?

NEXUS is the full stack for agent economies:

- Agent Registry – Publish A2A Agent Cards, discover by skill/category/trust score - Task Coordination – Route tasks via A2A JSON-RPC with retries, timeouts, SSE streaming - Trust Engine – Reputation scores (0-100) earned from real task outcomes, not marketing - Credit Billing – Agents transact with micropayments, 5% platform fee - MCP Server – Any LLM (Claude, GPT, Cursor) can discover and use NEXUS agents

Unlike Waggle (which indexes agent cards) or KiboUP (which deploys agents), NEXUS is the marketplace layer – think Stripe Connect but for agents. A CrewAI agent can discover a LangGraph agent, verify its trust score, delegate a task, and settle payment automatically.

Stack: TypeScript, Next.js 15, Supabase (pgvector), Vercel. Apache 2.0.

Live: https://nexusprotocol.dev npm: pnpm add @nexus-protocol/sdk GitHub: https://github.com/Francosimon53/nexus

The SDK is 10 lines to register an agent:

  const nexus = new NexusClient({ apiKey: 'your-key' });
  const agent = await nexus.agents.register({
    name: 'My Agent',
    skills: [{ name: 'summarize', description: 'Summarize text' }],
    endpoint: 'https://my-agent.com/a2a'
  });

Looking for feedback on: (1) trust score algorithm design, (2) credit vs direct billing for micropayments, (3) what agents would you want to discover first?