Product Manager · Berlin

Hi, I'm Ji Hyun — I solve problems that are misdiagnosed.

Product Manager with 4+ years in regulated, high-stakes products — AI, pricing, compliance. I bring a researcher's instinct to product: the stated problem is rarely the real one.

Get in touchSee experience ↓
Ji Hyun Kim
FocusRegulated B2B/B2C platforms, discovery, pricing, experimentation, AI features
Builds withReact, Next.js, Supabase, Vercel
LanguagesGerman (C2), English (C1), Korean (C1)
StudiedMSc Behavioral Economics, Erasmus University Rotterdam
CASE STUDY 01

Repricing a market leader

+30% YoY revenue+15% ARPU

Every property sold or rented in Germany needs a legal energy performance certificate. EA48 is the platform used to order one — serving both property owners and real-estate professionals.

EA48 charged the same price for a 3-unit townhouse as for a 20-unit apartment block, even though the operational work was wildly different. The pricing model couldn't defend ARPU as the market got more price-sensitive.

PROBLEMService tiers weren't differentiated — 'video' felt like a watered-down 'full service' rather than a real alternative. No sharp value story at the decision moment.
WHAT I DIDRebuilt pricing around building complexity (1–5 units vs 5+): larger buildings paid for the manual work they required, protecting ARPU. Repositioned video service on its own terms. Ran a sustained A/B programme on the landing pages — USP, tier positioning, pricing display.
SHIPPEDA new pricing architecture segmented by building size, a sharpened value story for each tier, and a redesigned landing page system.
CASE STUDY 02

AI for energy advisors

−18% issuance time

Before issuing a certificate, advisors have to check every data point for plausibility — historically a manual sweep of the full dataset.

EA48 held 200K past certificates — enough history to define what plausible values look like. The opportunity: auto-flag anomalies so advisors only reviewed what actually needed judgment.

PROBLEMA wall-to-wall sweep of every dataset, but most values were fine. The signal was buried in hours of routine review.
WHAT I DIDScoped the AI narrowly: compare each issuance against those 200K records, flag what looks off, suggest next steps — never the final call. The advisor reviews a shortlist, not the whole dataset.
SHIPPEDAn AI plausibility module in the issuance tool: compares values against historical records, flags anomalies, surfaces a suggested next step.

ALSO SHIPPED

CASE STUDY 03

DIN V 18599

A 2024 revision of DIN V 18599 — the German calculation standard behind every energy certificate — forced the whole market to re-architect on a legal deadline.

−60% turnaround time

Owned the product response end to end and automated processing — turnaround cut by 60%.

CASE STUDY 04

Vermietet.de research

ImmoScout24's property-management product for private landlords. I owned research across the landlord experience.

+13% satisfaction score

Ran interviews, usability tests, and Business × Tech alignment workshops.

Teams I've worked with

Where I come from

I run product like a researcher. That means talking to users before the spec, not after; writing problem statements, not feature lists; and treating the obvious interpretation as a hypothesis.

I started as a UX researcher at Vermietet.de, running discovery for several squads at once — interviews, usability tests, workshops that put Business and Tech in the same room. It taught me to find the problem under the ask.

Since 2024 I've been the only PM on Energieausweis48, working directly with a small engineering team on regulation, pricing, AI and reliability. The research instinct hasn't left — it just moved upstream of the roadmap.

Before that, I studied business and economics in Frankfurt, then behavioral economics in Rotterdam — because I kept wondering why people decide the way they do.

Nights and weekends

When I'm not talking to users, I'm probably sewing new clothes, making kimchi, or trying a new USC class somewhere in Berlin.

My notes app is full of ideas that never got built. Since I started working with Claude Code as my coding partner, a few of them actually shipped. I take each one from scoping to launch myself.

[01]
USC Workout Map

A PWA for coordinating Urban Sports Club sessions across a friend group. Our WhatsApp couldn't answer "who's at which studio tonight?" — a coordination gap, not a feature gap.

Validated with the friend group as a research proxy before building. Scoped ruthlessly: map, group sessions, push reminders. Shipped iteratively, pushed to prod the moment each milestone was usable.

ReactSupabaseMapboxWeb Push
shipped ↗
[02]
This portfolio

Scoped from a design handoff, shipped in a weekend. Said no to a CMS, a blog, and animations on scroll — only built what served the brief.

Next.jsVercel
2026

What I've shipped

Open full CV ↗Download PDF ↓
Mar 2024 — now
Berlin

Product Manager

ImmoScout24 / Energieausweis48 · sole PM

  • Discovery & experimentation: built a feature from customer pain points (+11% conversion); owned A/B tests end to end
  • Scoped and prioritized B2B API integrations with ImmoScout24
  • Reliability: led refactoring with engineering, reducing error rates by 19%
  • Data Protection Coordinator for GDPR-compliant product decisions
Apr 2022 — Mar 2024
Berlin

UX Researcher (Junior → Professional)

ImmoScout24 / Vermietet.de

  • Competitive benchmarking that shaped strategic roadmap priorities
  • Supported the Vermietet.de integration into ImmoScout24 via Business × Tech alignment workshops
Earlier

Founder's Associate, T-rtle · Business Development, Tomorrow · UX Research, DKB Code Factory

Education
MSc Behavioral Economics · Erasmus University Rotterdam2020 — 2021
BSc Business Administration & Economics · Goethe University Frankfurt2016 — 2019

Let's build something meaningful.

jihyunkim749@gmail.com
LinkedIn ↗+49 1573 9081140