ivdim

We find the way into your cloud and your AI agents before someone else does.

Cloud and AI security for startups and small teams. We map how an attacker would move through your AWS accounts and the agents you've connected to them, then help you close the gaps that matter most.

Two paths we close in a typical first week. Each starts somewhere public and ends somewhere it shouldn't. One change breaks it.
  1. Internet
  2. Public load balancer
  3. EC2 instance
  4. Scope the instance role
  5. Instance role
  6. Account admin
  1. Email with hidden instructions
  2. Support agent
  3. Make the tool read-only
  4. Database MCP tool
  5. Customer records

Services

AWS security assessment

A read-only review that traces real attack paths through your accounts, from public endpoints to admin rights and data, and ends with a short, ranked fix plan.

AI and agent security review

An inventory of your models, agents, MCP servers and API keys, a threat model for prompt injection and tool abuse, and least-privilege scopes for each agent.

AI-assisted security operations

VulShark deployed on your estate, with specialist AI agents that investigate risk and record findings with evidence. Runs on Claude, Amazon Bedrock or a local model.

Identity and access

Single sign-on with SAML and OIDC, IAM Identity Center, MFA gaps and least-privilege roles.

Secure cloud foundations

AWS Organizations and Control Tower landing zones, Terraform, and hardened CI/CD pipelines.

Ongoing security cover

A monthly retainer for teams without a security hire: monitoring, design reviews and help when something looks wrong.

Things we've built

VulShark

A self-hosted cloud and AI security platform. It maps AWS accounts and connected tools into one graph, traces attack paths to admin rights, data and AI, and runs specialist agents that record evidence-backed findings.

Tell us what you're running

Send a line about your stack and what's worrying you.

me@ivdim.cc