Hey — you probably got an email from me.

I'm Issac. I build AI governance tools that let regulated teams ship LLM features without getting sued, fined, or owned. Three products, three real prices, no sales deck.

This page exists so you can decide on your own terms. No calendar booking pressure, no gated PDFs, no "request a quote." Here's what's actually available:

If you're wondering why you should trust me

I've been building this for a while. The framework underneath all three products is open source (MIT), patent pending, and ships with 99.42% combined AUC and 91/91 red team prompts blocked in CI. It's not a deck. It's running code you can clone.

Hey — you got an email from me about chatbot liability.

Quick version: after Moffatt v. Air Canada, any company running an AI chatbot is legally on the hook for what it promises. One wrong answer = a binding contract. I built CX Refund Guardrail to prevent that, specifically for mid-market teams who can't afford Decagon or Sierra.

What it actually does

It's policy-enforcement middleware that sits between your LLM and your customer. Every outgoing message gets scored against your actual policies before it ships. Anything outside your allowed envelope gets caught, redirected, or escalated to a human.

• Works with any LLM (OpenAI, Anthropic, local, custom fine-tunes)
• Under 100ms added latency per message
• Policies live in YAML or JSON, versioned in your repo
• Audit log of every decision, ready for legal/compliance
• 30-day trial, up to 1,000 messages, no credit card

What it costs

Full details, included features, and FAQ on the CX Guardrail page.

Hey — you got an email from me about AI audit readiness.

Three regulatory frameworks are converging on your AI deployments in 2026: ISO 42001, SR 11-7, and the EU AI Act. Your auditors are going to start asking for documented adversarial testing, risk assessments, and drift monitoring. Most banks and insurers already deployed LLMs and don't have that paper trail. I deliver it.

What I actually deliver

• Test evidence package — JSONL output of 91 red team prompts (0 false positives) run against your LLM, plus 150/150 compliance tests across 13 frameworks, mapped to ISO 42001 clauses
• Risk assessment report — formatted for SR 11-7 review, signed and timestamped
• Drift monitoring setup — ongoing adversarial probe suite with quarterly re-tests
• Audit response dossier — everything your compliance team hands to auditors, pre-formatted

I don't certify you — certification is a separate process. I deliver the evidence your compliance team presents to auditors. You ingest the output into Vanta, Drata, or whatever GRC platform you already use.

What it costs

Full scope, ISO clause mapping, and FAQ on the ISO 42001 Evidence page.

Hey — you got an email from me about stress-testing your LLM.

If you've shipped an LLM feature — a chatbot, an agent, a RAG system — you need to know how it breaks before an attacker or a reporter does. I run the 91-prompt red team (0 false positives) against your application, deliver a branded PDF report mapped to your threat model, and hand you a prioritized remediation roadmap. Four weeks. Real findings.

What we test for

• Prompt injection — direct, indirect, multi-turn, context poisoning
• Jailbreaks — persona hijacks, instruction override, safety classifier bypass
• Data exfiltration — system prompt extraction, training data leak, PII recovery
• Agent abuse — tool misuse, privilege escalation, infinite loops, cost bombs

What the engagement looks like

Week 1 scoping call → Week 2 execution against your staging endpoint → Week 3 analysis and triage → Week 4 delivery call with your engineering and security team. Zero impact on production. NDA signed before any API keys change hands.

What it costs

Full scope, example findings, and FAQ on the Red Team page.