01 · B2C
Career Companion
A quiet co-pilot for the job hunt.
Document analysis, tailored cover letters, and AI mock interviews — the candidate side of the platform, designed to feel calm rather than transactional.
The Patch
A complete AI recruiting ecosystem built from zero to one. A B2C career companion (document analysis, writing, mock interviews), a B2B evaluation platform that screens candidates on measurable skills and AI interviews, and a school dashboard that helps institutions track and support their students.
- Role
- Product Manager & Designer
- Year
- 2025 — Present
- Ingredients
- Figma · Lovable · Notion · OpenAI · ElevenLabs · Mixpanel · Claude Code
- Price
- €28
Three surfaces, one product
An ecosystem for AI recruiting
The Patch is built around three connected audiences. Each surface ships as its own product, but they share a single data spine — what a candidate practises on the B2C side feeds the skill profile a recruiter sees, and the school dashboard sits above both.
02 · B2B
Skill-based Evaluation
Hiring decisions grounded in skills, not keywords.
An evaluation platform that screens candidates against measurable skills through structured AI interviews — surfacing depth recruiters can't get from a CV scan.
03 · EDU
School Dashboard
Track and support students at scale.
Helps institutions monitor candidate progress, intervene early, and report on outcomes — closing the loop between training and employment.
How it was built
From wireframe to live product
I owned the loop end to end — design, specification, prototyping, instrumentation. The stack stayed small on purpose, so shipping a new feature could happen in days rather than sprints.
- 01WireframeLovable
Used Lovable to sketch interactive wireframes for every feature — fast enough to share a working flow inside the same week the idea landed.
- 02Hi-fi prototypeFigma
Partnered with a designer in Figma to push the wireframes into polished, production-ready screens — building the visual system alongside it.
- 03SpecsClaude Code · Notion
Wrote detailed PRDs that engineering and Claude Code could pick up directly — typed flows, edge cases, copy decks, and acceptance criteria, all version-controlled in Notion.
- 04VoiceElevenLabs
Prototyped the mock-interview voice agent on ElevenLabs — testing latency, tone, and interruption handling before committing to a production voice.
- 05EvaluationOpenAI
Built the post-interview analysis pipeline on OpenAI — scoring transcripts against rubrics and producing the feedback candidates actually see.
- 06InsightMixpanel
Instrumented every surface in Mixpanel from day one — funnels, retention cohorts, feature-level adoption — and made weekly product decisions from the data.
The bar cart
Toolkit
Design
Figma · Lovable
Specs
Notion · Claude Code
AI
OpenAI · ElevenLabs
Analytics
Mixpanel
Methods
PRD writing · Prototyping · Beta interviews · A/B
What it produced
Outcome
6k
Users on the B2C product
3
Paying B2B clients
3
Key-Account PoCs shipped
96%
Reduction in client-specific dev time
The 96% figure comes from a modular PoC framework I built so a new client demo could be assembled from configurable building blocks instead of a fresh codebase.
Looking back
Reflection
Own the loop
Doing design, PRDs, and analytics together meant I could close the gap between an idea and a measurable result in days — and reroute when the numbers disagreed with my intuition.
Tools as collaborators
Lovable and Claude Code shifted what a single product person can ship. Specs became running prototypes; prototypes became released features.
Calm UX in a noisy category
AI products tend to feel loud. The Patch leans into restraint — fewer surfaces, clearer language, and honest feedback over flashy outputs.