Work AI Builds Practice About Contact
Available full-time · Remote, hybrid or onsite · Open to relocation
Senior Product Designer

Product design for
AI and data systems
people actually trust.

I design products where the hard part is the data — live sensor feeds, maps, time-series, alerting rules, regulation. 10+ years of it: betting platforms under strict compliance, hotel operations software with brutally complex setup flows, and my own real-time monitoring platform, now the leading service of its kind in Serbia. I also build, which changes how I design.

10+
Years in product design
40+
Enterprise workflows rebuilt
1,585+
Live stations designed for
15k
Monthly users on my own product
01 — Case Studies

Selected work

01
AI & real-time dataZero-to-oneRule engineFounder

Vodostaj.rs

Founder & Product Designer · Solvers Hub DOO (solve.rs) · Jan 2026 – Present

Real-time monitoring platform built from zero: 1,585+ hydrological stations across six countries, fed by automated scraping and API integrations with national meteorological agencies. Map interface, time-series views, and the threshold model behind flood and drought alerts — every alert traceable to the rule that fired it.

15,000 monthly users1,585+ live stations6 countries89% mobile
02
Flagship enterprise caseEnterprise SaaSEnd-to-endDesign system

Hotel Sync

ex OTA Sync · Aug 2022 – Jan 2024

End-to-end redesign of an enterprise operations platform from 1.0 to 2.0 — the PMS back-office, a native manager app, and the guest-facing Booking Engine. Re-architected 40+ configuration screens, defined the pattern library engineering built against, and stayed embedded through the rollout.

25–60% fewer clicks−15% support tickets+20% onboarding3 surfaces owned
03
RegulatedA/B testingConversionShips code

Bethard / Uberlucky

Betting Platform · Feb 2021 – Aug 2022

Hypothesis-driven conversion work and end-to-end product design on a licensed betting platform, under compliance constraints on every screen. Designed the full Uberlucky web product — homepage, KYC-heavy registration funnel, account panel, deposit and bonus systems — and delivered production CSS for the sportsbook module myself.

+8% deposit click upliftFull product designedProduction CSS delivered

Supporting work

B2B SaaSOnboarding

Unternehmerheld

B2B SaaS, Berlin · Aug 2022 – Jan 2024

Registration and onboarding redesign for a German B2B financial planning SaaS. Worked directly with PM and CEO, ~6,000 active users.

● Current roleResearchUsability testingAnalytics

Maxbet

Part of Flutter Entertainment Group

Feb 2025 – Present

Stood up the research and insight practice for a 28-person product org that had none: GA4 event taxonomy, FullStory funnel analysis, Maze usability validation, Looker health dashboards, and a shared research repository — all inside a compliance-heavy regulated environment.

Zero-to-oneWeb designSide project

Luigi Rent a Boat Hvar

rentaboat-hvar.com · Side project

End-to-end web design for a Croatian boat rental and tours business. Homepage, boat rental pages, tour listings, water taxi, and full mobile responsive — built and live.

Zero-to-oneAI-designedSide project

AKC Gnezdo

akcgnezdo.rs · Side project

End-to-end site designed entirely through Claude — no Figma. Public homepage, event pages, calendar, gallery, and a full admin panel for the editor to run everything. Built for friends, also a proof that AI-assisted design can deliver a production site with full management tooling, fast.

02 — AI Product Design

Designing for AI

Most of my recent work sits on top of automated output — rule engines, threshold evaluation, forecasting, behavioural models. The interesting design problem is never the model. It's the moment a person has to decide whether to act on what the system just told them. These are the principles I design to.

01

Traceable, not oracular

An automated flag is only useful if the user can see the rule behind it. Every generated output I design links back to its source — which threshold, which regulation, which reading, at what timestamp. A verdict nobody can audit is a verdict nobody acts on.

02

Confidence is part of the data

A forecast shown without an error bar reads as a fact. I design predictive UI to expose its own uncertainty — range, data freshness, and the conditions under which it degrades — so people calibrate their trust instead of inheriting it.

03

The human keeps the decision

The system narrows the field; the person still chooses. Review, override, and disagreement are first-class states in the flow, designed up front — not error handling bolted on after the model ships.

04

Design the failure case first

Stale feeds, dropped sources, low-confidence output, zero results. In a live data product these are not edge cases, they are Tuesday. I specify the degraded states before the happy path, because they decide whether people keep the product open.

05

Complexity belongs in the system

Dense domains earn their complexity — hydrology, hotel operations, betting compliance. The work is to absorb it into structure, hierarchy, and progressive disclosure so the surface stays operable by someone who is not an expert in it.

06

Prototype it, don't argue about it

I build working prototypes with AI-assisted development, so a contested flow becomes something a team can click through in days rather than something they debate over a static mock for weeks.

Where this is applied
03 — Selected Builds

Designed it, then shipped it

Products I designed and built myself, end to end, through Solvers Hub. Evidence of implementation awareness, full-stack product thinking, and the technical acumen to prototype an idea rather than only specify it. Running in production on Next.js, TypeScript and Supabase — which is why I know what is cheap to build, what is expensive, and what breaks.

My studio · Solvers Hub DOO · Next.js
Solve.rs ↗

The product studio I run alongside my primary role — everything shipped end to end by me. Vodostaj.rs was migrated from a React SPA to Next.js here and now sits past 2.6M impressions and 60K organic clicks at #1 regional rankings. Design decisions measured against outcomes, not opinion.

Next.jsTypeScriptSupabaseClaude
Geospatial · MapLibre · Crowdsourced
Environmental mapping platform

Crowdsourced environmental reporting for the Balkans. Full geospatial pin lifecycle on MapLibre and OpenStreetMap — submission, EXIF-aware image pipeline, moderation admin, and resolution — across six languages. The closest thing here to a field-reporting and inspection tool.

MapLibreOpenStreetMapEXIF pipelinei18n ×6
E-commerce · Next.js · Stripe
JigBau.com ↗

Full-stack e-commerce for a fishing tackle brand: catalogue, Stripe checkout, order management, and an admin dashboard the client runs themselves. Empty repo to production in a 24-hour sprint.

Next.jsStripeAdmin dashboard24h sprint
Multi-tenant · Institutional
Regional sports federation platform

Multi-tenant platform for a regional sports federation — 85 clubs across 40+ disciplines, each with its own content, membership and permissions, under one administrable roof. Plus client platforms for manufacturing, hospitality, and multi-location service businesses.

Multi-tenantSupabase RLSAdmin CMS85 tenants
04 — Practice

What I bring, and how I work

Time-series & data visualisation

Charts with thresholds, min/avg/max bands and confidence ranges; occupancy and revenue dashboards; funnel analytics. Making dense data readable at a glance is the core of most of my work.

time-series · thresholds · KPI dashboards

Geospatial & mapping interfaces

Station networks and pin lifecycles on MapLibre and OpenStreetMap, with EXIF-aware image pipelines and moderation tooling behind them. Designing the map is easy; designing what happens at 1,585 pins is the work.

MapLibre · OpenStreetMap · GIS · clustering

Threshold, alerting & risk detection

The rule model behind flood and drought alerts: how a user defines a threshold, how the system explains which rule fired, and how you earn the right to interrupt someone. 134 user-set alerts, all still enabled.

rule modelling · traceability · notification UX

Enterprise workflow architecture

40+ configuration screens regrouped into domain-based navigation. A nine-step reservation flow reduced to four. The value in a complex product is decided in its structure, long before anything is styled.

IA · progressive disclosure · state modelling

Design systems & foundations

Twelve inconsistent interaction patterns consolidated into one modular component system and pattern library engineering built directly against. Comfortable being the person who establishes design foundations where none exist.

Figma · component systems · pattern libraries

Research & usability validation

Discovery interviews, event taxonomies, session and funnel analysis, moderated and unmoderated usability testing, competitor benchmarking — plus the repository standards that keep findings usable after the study ends.

GA4 · FullStory · Maze · Looker · Confluence · Miro

Compliance-aware UX

Licensed betting at Bethard and Maxbet, German financial planning at Unternehmerheld, public-safety-adjacent hydrology at Vodostaj. Translating a rulebook into something people can actually use is half the job.

regulated markets · auditability · disclosure

Institutional data integration

Automated scraping and direct API integration with national meteorological agencies across six countries, behind a source-adapter layer. Each source has its own format, cadence, and way of failing — the interface has to absorb all of it.

government APIs · ingestion · multi-tenant

Prototyping in code

Next.js, React, TypeScript, Supabase, Tailwind, Stripe. I build what I design when building it is the faster way to settle a question — and it means I know what is cheap to build, what is expensive, and what breaks.

Next.js · React · TypeScript · Supabase · Stripe

How an engagement runs

STEP 01

Frame

Get to the real problem with the people who own it. Audit the existing system, map every user task to the steps it actually costs, and agree what "better" means in numbers before anything is designed.

STEP 02

Architect

Structure first — information architecture, states, edge cases, and the rules the system runs on. This is where most of the outcome is decided, and where I spend the most time.

STEP 03

Prototype & validate

High fidelity in Figma, working prototypes in code when that answers faster. Test with real users, instrument the flow, and let behaviour settle the disagreements instead of seniority.

STEP 04

Ship & measure

Stay inside the implementation cycle — review builds, adjust in flight, and tie the change back to click cost, support load, activation, conversion. A design isn't finished at handover.

Senior Product Designer, 10+ years across enterprise SaaS, licensed betting, embedded engineering, and real-time data platforms. Systems thinker, comfortable with ambiguity, used to being the only designer in the room and the one who establishes the foundations. Based in Serbia (CET) — available full-time, remote contract or permanent, and open to relocation.

Download CV →
Beyond the screen

„Prozor u prošlost" — A Window to the Past. A public heritage installation at the Kruševac fortress: a glass panel that visually completes the ruined medieval Donžon tower when you stand in front of it, drawing the missing structure back in at full scale, in place. It has gone viral several times and is now a small landmark of my hometown. Same instinct as everything else on this page — take a thing people cannot read, and make it legible.

Lifelong angler, which is where Vodostaj.rs came from: I built the tool I needed on the riverbank. I like products born from real, personal problems.

Let's talk about
your product.