People-centered AI products and consultancy. We ship the lab's own product line and the same team delivers custom builds for your business — never one at the cost of the other.
Voice agents, GCP builds, custom RAG, CI/CD pipelines, and multi-agent architecture — deployed by our team, owned by yours. Our product line (SOS.VU, cam.vu, CALEL Legion, Aquaman) funds the lab and stress-tests every technique before we bring it to you.
Voice-first · typed fallback if you prefer. Ask about the four consultancy pillars, any of the 13 products, or how we work — the agent can book you a discovery call or open the intake form when you’re ready.
Under our AI Overlords partnership we implement Voice AI (ElevenLabs, Vapi, Retell, Bland, Deepgram, Cartesia, Speechmatics), GCP AI-platform builds, custom RAG stacks, CI/CD pipelines, and multi-agent architecture. Commission-based reseller-plus-implementation model, sold as working product not as hours.
Three snapshots from recent implementations. Names blurred where confidentiality is still active. Full case studies open up after a scoping call.
Turn-taking sub-350ms. Warm hand-off to human when the agent's confidence dropped. Booking conversion up meaningfully in the first month. ElevenLabs voice model plus Vapi orchestration plus custom evals.
Five languages, one agent, latency budget hit. Domain-specific RAG for property information. Fall-back to human staff when the guest asked for human contact. Deployed in weeks.
A pilot for personal safety in Warwickshire. Native app with web backup for un-signed users. Testing phase live, public onboarding held back until we are happy. Privacy-scrubbed metrics on request.
Fifteen prior engagements across financial services, telco, pharma, consulting and startups — direct by the founder or via the AI Overlords consultancy arm. Framing stays historical.
Voice AI. GCP. Custom agents. RAG. CI/CD. Shipped on-site or cloud — you own the code.
What we are shipping, what we are learning, what we would build differently. Written from the desk, not the deck.
Turn-taking sub-350ms is table stakes. Warm human hand-off is table stakes. What makes the difference between a demo and a working production agent is the eval loop — and how tight the scope was on day one.
Vertex AI, Cloud Run, BigQuery, Firestore, Pub/Sub. Where each earns its cost line and where it is a tax you pay for convenience. Coming soon.
Handover ladders, state monitors, and how to stop drift between agents built on different LLMs. Coming soon.
If you have a real problem and a real timeframe, tell us. We reply within two working days in London. If we can help, you get a working session. If we cannot, we say so and point you somewhere better.
Open the intake form