I’m a builder, so I’d rather show you the work. Below is a working prototype I’ve built, followed by concepts I’m designing. I’ve labelled which is which.
Screening call finished. Report is in the inbox.
045°
Placeholder · replace with a real photo of water, a chart or a compass
Recruitment · Working prototype, not deployed with a client
AI recruitment integration
The problem
Recruiters spend a large part of their week on first-round phone screens, and searching a big candidate database by keyword misses good people who describe their experience differently.
AI voice screening interviews
Candidates receive an AI phone interview of around 10 to 15 minutes. The voice agent asks structured screening questions and has a natural conversation. When the call ends, an automation scores the interview against a rubric and sends the recruiter a clear, formatted report in Outlook, ready to review.
Built with Retell AI and Twilio for the voice and phone side, and n8n for the scoring and reporting workflow.
Semantic CV search
A search tool that finds candidates by meaning, not just keywords, so “managed a busy warehouse team” can match a search for “logistics supervisor”. It includes filters for location and salary expectations. I’ve tested it on development data sets, and it’s designed to handle a database of around 400,000 candidates. It hasn’t been deployed at that scale.
Built with OpenAI embeddings and Pinecone.
What I learned about integrations
I worked on connecting the prototype to an applicant tracking system, and the integration stalled because the vendor wanted to own the AI features inside its own product. It’s a useful lesson: before building, I check that the systems you rely on will actually let you connect to them.
Why it matters for other businesses
The same pattern works well beyond recruitment: a voice agent handles a structured first conversation, AI turns it into a consistent summary, and a person makes the decision. That applies to enquiries, bookings, intake and qualification in many industries.
In design
Concepts in design
These are designs and prototypes in progress, not live products with clients. I’m sharing them because they show how I think about problems in different industries.
Concept
Cloud Waiter
A voice agent that takes phone orders for restaurants and takeaways, answers menu questions and sends orders through to the kitchen, so staff aren’t pulled away from customers in the room during the busiest hours.
Concept
Real estate intake agent
An agent that handles the first conversation with new enquiries (buyers, tenants or vendors), gathers the key details, qualifies them and books the next step with the right person.
Concept
Voice desktop assistant
An assistant you talk to while you work, to draft emails, find information, update records and handle small admin tasks without breaking your flow.