
Tom Phillips
AI Solutions Architect. Cybersecurity & Applied AI.
I build practical AI systems that solve real problems in governance, risk, compliance, and automation — from commercial SaaS platforms to intelligent agents that eliminate manual processes.
Tom Phillips Labs is the working name for it — applied-AI experiments, small tools and shipped products, each one built in the open and written up.
What's running
- OpenAI GPT-4o
- Azure OpenAI
- Anthropic Claude
- Whisper
- RAG
- Dify
- n8n
- Vector DB
- ElevenLabs
- LangChain
- Python
- FastAPI
- Supabase
- PostgreSQL
- Stripe
- Next.js 16
- React 19
- TypeScript
- Tailwind v4
- Capacitor
- Vercel
- Azure
- Docker
- RSA Archer
- OneTrust
- ISO/IEC 42001
- EU AI Act
- RBAC + RLS
Flagship projects
Production systems with interactive sandboxes.
Smaller tools
Built to scratch a real itch. Live and free.
SentinelProxy
Self-healing for AI agents.
An asynchronous log-ingestion platform that watches AI agents in production, catches the hallucinations and prompt injections a firewall only blocks, and hands back a patched system prompt. The self-healing layer for Dify, n8n and LiteLLM.
- Webhook ingest returns 200 instantly — never on your critical path
- PII scrubbed at the edge before anything is stored (UK GDPR)
- A background judge grades every call against your guardrails
Next.js · TypeScript · Supabase · Claude Haiku 4.5 · Vercel
Delivered for enterprise teams
Anonymised case studies — applied AI inside live enterprise environments. Hover to pause, drag to browse.