Scheduling one meeting took 39 emails. Now it takes none.
An AI assistant that arranges small-group meetings entirely over email. No app, no login, no account — for the organiser or any guest.
60–80% less scheduling time, per TroikaBot
// case studies
Each of these is a product we delivered end to end, not a feature bolted onto someone else's codebase. The stack is listed because you'll want to know what you're inheriting.
An AI assistant that arranges small-group meetings entirely over email. No app, no login, no account — for the organiser or any guest.
60–80% less scheduling time, per TroikaBot
Matching Swiss homeowners with vetted tradespeople nationwide, with language treated as a matching constraint rather than a UI toggle.
4 languages, 11 trades, nationwide
Medication tracking, macros, weight and meal planning in a single app for people on GLP-1 medication. Logging a meal costs one tap.
1 tap to log a meal, on two platforms
Voice agents that qualify inbound callers on budget, timeline and area, write the lead into GoHighLevel live, and hand the serious ones to a human.
0 missed enquiries, day or night
// in-house builds
Products with no client attached, which means we can show you the source. Same engineering, fewer things we're not allowed to say.
Question a PDF and reach its charts and tables, not just its prose — with the page the answer came from attached. Live, public, and yours to try.
3 content types, one index — text, tables, charts
Route to one rule category first, fetch by ID second. Similarity search never gets the chance to return a statute that merely sounded right.
1 model call per document — the rest is embeddings
Ranked paths and skill gaps derived from real job descriptions, so the roadmap points at roles that are actually open.
4 outputs from one upload: paths, gaps, roadmap, openings
// also shipped
Smaller builds and systems that did not warrant a full write-up, listed with the stack so you can see the range.
05
A RAG platform that ingests lecture PDFs and audio recordings and assembles a personalised learning path from them, cutting the time a student spends working out what to study before they can start studying.
Jun 2025 – Jun 2026
06
An AI-assisted publishing platform with LLM content-review agents and RAG pipelines behind the editor, built around writer monetisation rather than raw content volume.
Mar 2026 – Jun 2026
07
An NLP recommendation engine for an e-commerce platform, reading customer behaviour and purchase history to personalise what each shopper is shown.
Sep 2025 – Dec 2025
// start here
Tell us the number you're unhappy about. We'll tell you on the call whether we think we can move it, and roughly what it takes.
Book a call