Selected Work & Projects

Key case studies across my professional responsibilities at LG Electronics Benelux and academic projects at the University of Amsterdam.

LG Electronics Benelux

01 / PROFESSIONAL EXPERIENCE

LG TV & Audio Launches

Featured Rollouts Assistant Product Manager

Launching LG TV & Audio Products Across Benelux

Coordinating end-to-end commercial launch workflows, cross-functional dependencies, and market execution for TV & Audio lineups (including the April 2026 Sound Suite launch) across NL & BE.

Commercial Reporting Transformation

Process Automation Reporting & Analytics

Commercial Reporting Transformation

Migrated commercial workflows to BigQuery/OLAP, reducing report prep time from 60 minutes to under 5 minutes.

Power BI Dashboard

Business Analytics Business Analysis & BI

Power BI Dashboard Requirement & Validation

Partnered with data engineering to define requirements and validate 3 dashboards used daily by 10+ Sales stakeholders.

University of Amsterdam

02 / ACADEMIC & LAB PROJECTS (MSc)

Machine Learning

Data Science & AI CRISP-DM Framework

Predicting Diabetes Risk with Machine Learning

Built an end-to-end CRISP-DM predictive model analyzing 253,680 records to identify high-risk health profiles (0.822 Test Set AUC).

B&T Decision Tool

Agile Project Lead B&T Educational Consultancy

Building a Data & AI Decision-Support Tool

Led an Agile team delivering an AFAS-connected Power BI dashboard integrated with localized Microsoft Copilot AI for executive decision-making.

Annual Rollouts

Product Lifecycle

2 Markets

Netherlands & Belgium

4 Workstreams

Sales / HQ / Marketing

Launch Architecture & Commercial Workflow

HQ Alignment

→

Benelux Roadmap

→

Partner Enablement

→

Retail Launch

→

Sales & GfK Analysis

Key Execution Pillars

Cross-Functional Timeline Management: Aligning local product launch milestones across Product, Sales, Marketing, and E-Commerce teams to hit synchronized commercial launch deadlines across Benelux retail channels (including key milestones like the Sound Suite launch).
Partner & Sales Enablement: Translating HQ technical specs and promotional guidelines into channel-ready collateral, product cheat sheets, and digital asset packages for key Benelux retailers.
In-Store & Online Execution: Monitoring physical display installations at key retail partners while verifying PDP (Product Detail Page) accuracy and bundle offerings across e-commerce platforms.

Post-Launch Category Operations & Market Logic

Market Intelligence (GfK): Continuous benchmarking of sell-out rates, average selling price (ASP), and market share dynamics against primary competitors in TV and soundbar segments.
Inventory & PSI Management: Monitored channel inventory buffers and stock turnover rates to prevent stockouts during promotional campaign peaks.

Commercial Impact: Bridged high-level global strategy with granular regional execution, ensuring full channel readiness, accurate product positioning, and proactive post-launch sales optimization across Benelux trade partners.

Go-To-Market Strategy TV & Audio Lineups Sound Suite Launch GfK Market Analytics Retail Execution Cross-Functional Alignment

253,680

Health Records

0.822

Hold-out Test AUC

83.69%

Risk Recall Rate

21 → 10

Feature Selection

Data Pipeline & Imbalance Strategy

Class Imbalance Handling: Dataset contained 86.1% majority (No Diabetes) vs 13.9% target (Diabetes). To prevent accuracy paradox, majority class downsampling was applied strictly within training folds during 5-fold cross-validation. Validation/Test sets retained natural distributions.
Feature Selection: Information gain ranking and correlation filters reduced 21 raw variables to 10 primary indicators: General Health, High BP, Age, Difficulty Walking, High Cholesterol, BMI, Income, Heart Disease, Physical Health, and Education.

Model Benchmark & Validation Matrix

Algorithm	Model Type	Cross-Val AUC	Test Set AUC
Gradient Boosted Trees ★	Boosted Ensemble	0.826	0.822
Logistic Regression	Linear Baseline	0.822	0.819
Random Forest	Bagged Ensemble	0.821	0.817
Decision Tree	Rule-based Tree	0.798	0.791

Business Trade-off & Decision Logic

Prioritizing Recall Over Precision: In health risk screening, a False Negative (missing a diabetic patient) has severe health and financial consequences. The model probability threshold was tuned to achieve 83.69% Recall, ensuring the vast majority of high-risk cases enter preventative intervention funnels.
Generalization Stability: Minimal AUC delta between Cross-Validation (0.826) and Hold-out Testing (0.822) confirms robust generalization without overfitting.

RapidMiner CRISP-DM Gradient Boosted Trees Feature Selection ROC/AUC Optimization

Feb–May 2026

Project Timeline

5 Members

Cross-Functional Team

3-Tier

AFAS → Power BI → AI

System Architecture & Solution

AFAS ERP Data

→

Power BI Dashboard

→

Localized Copilot AI

→

Management Insights

Agile Project Management Responsibilities

Roadmap & Backlog Ownership: Led sprint planning, story estimation, and backlog prioritization using Jira & Confluence to maintain project momentum across 2-week development cycles.
Scope Safeguarding (MoSCoW): Applied strict MoSCoW prioritization to protect core MVP deliverables (Consultant Capacity, Billability Rates, Project Financials) against non-essential feature requests.
Stakeholder Alignment: Managed communication between executive leadership and technical engineers to resolve conflicting definitions around pre-calculation (voorcalculatie) KPI logic.

Technical Challenges & Prompt Engineering

Copilot AI Dutch Localization: Microsoft Copilot struggled with native Dutch terminology due to English workspace configs. Engineered a 20-prompt bilingual validation matrix and aligned dataset metadata in Dutch to achieve >90% accuracy for direct conversational queries.
Dependency & Blockers Management: Proactively mitigated delays surrounding AFAS API permissions and Microsoft Copilot tenant licenses by re-sequencing frontend dashboard tasks during IT security reviews.

Outcome: Replaced fragmented manual Excel tracking with a unified live dashboard and AI natural-language interface, drastically improving consultant capacity visibility and executive decision velocity.

Jira / Confluence Agile / Scrum Power BI Microsoft Copilot Prompt Engineering AFAS Integration