Every product we ship is grounded in a problem our founders and their communities had — problems that are widespread and that we deeply feel need resolving. Personal safety, every day and at events. Beautiful diagrams and documents from your LLM chat, without wrestling with a design tool. Your family's moments in one place, not scattered across five apps.
The same team also builds AI systems for management consultancies, marketing agencies and founders adding AI capability quickly. Our product line (SOS.VU, cam.vu, CALEL Legion, Aquaman) is available exclusively to our clients before public launch and post stress-tests on every agent before we bring it to you.
If you're a management consultancy or marketing agency adding AI capability quickly, or a founder who needs a working prototype in weeks — this is what we build. Voice agents that answer your phone at 3am. GCP + RAG stacks that give your team its own knowledge back. Multi-agent architecture that doesn't drift. Delivered as products and implementations, not as hours. If your work involves lone workers or event safety, our own product line handles that — start with SOS.VU.
Three recent snapshots of customers who came with a problem and left with a working system. Names blurred where confidentiality is still active. Full case studies open up after a scoping call.
Their customers wanted to book at all hours. Their team wanted to sleep. A real-time voice agent answers now with sub-350ms turn-taking, hands off warmly to a human when it isn't sure, and booking conversion went up meaningfully in the first month. Built on ElevenLabs voice, Vapi orchestration, custom evals.
Their front desk couldn't cover every language every hour. Their guests wanted answers in their own words. One voice agent now speaks all five, pulls answers from a property-specific knowledge base, and hands off to human staff the moment a guest asks for one. Deployed in weeks.
People with real anxiety about walking home late, working alone, or living alone. SOS.VU is in private test right now — native app with web backup for anyone not signed in. Public onboarding is held back until we are happy every escalation ladder works the way a person under stress needs it to. Privacy-scrubbed metrics on request.
Fifteen prior engagements across financial services, telco, pharma, consulting and startups. The kind of problems those teams brought — and the kind we still take on today — range from voice agents on customer lines to knowledge systems that let a whole department search their own past work in plain English.
Bring us the problem your team has been putting off. You leave with a working system, on-site or in your cloud, and your team owns the code.
Notes from the workbench for anyone building AI into their business or their product. What actually works, what quietly falls over in production, and what we would do differently if we started again on Monday.
If your team is being told a working voice agent needs a six-month engagement, someone is padding. Here's what actually takes time, what doesn't, and what the eval loop looks like — so you can tell a real timeline from a scoped-to-bill one.
Why modern personal safety isn't solved by another wearable button that sits uncharged on a nightstand. What an ambient safety network looks like when latency actually matters.
The single biggest reason production multi-agent systems fall over is bad handovers between agents on different models. Handover ladders, state monitors and the guard-rails that catch drift before your customer sees it. Coming soon.
Tell us the problem your team keeps putting off, and the timeframe you're up against. We reply within two working days in London. If we can help you ship, you get a working session. If we cannot, we say so and point you to someone who can — you leave better-off either way.
Open the intake form