Making technical expertise accessible through intelligent search and semantic structure

Challenge

As one of Germany’s leading technical associations, the VDE manages a vast amount of specialist content — from standards and position papers to expert articles and research outputs. But the existing knowledge environment was fragmented and difficult to navigate. The challenge: create a smart platform that connects users with the right content across silos, formats, and departments.

Project Overview

We were part of the digital transformation team working on the conception and design of a central AI-powered knowledge platform for VDE. Our focus: enable intelligent search and intuitive access to expert knowledge by using semantic tagging, contextual relevance and smart clustering.

The system not only improves content findability — it becomes the backbone of a smarter internal and external knowledge experience.

Features

• Collaborative concept for AI-driven content indexing and semantic search
• User-centric interface logic for knowledge exploration
• Wireframes and navigation models tailored to content types and access levels
• Scalable backend structure for future content growth
• Cross-functional coordination between tech, legal, editorial and domain experts

Outcome

• Stronger accessibility of VDE knowledge across audiences
• Foundation for smart expert systems and internal reuse
• Improved UX for members, industry partners and staff
• Future-ready knowledge infrastructure with room for AI expansion

Expertise is only valuable when it’s accessible. We helped build the bridge between structure and meaning.

Want to unlock your organization’s knowledge with AI-powered tools?

Let’s explore how.

Image Recognition + Visual Data Clustering

Challenge

Understanding what kind of motorcycle imagery drives the highest user engagement on Instagram can be complex when dealing with 10,000+ visual posts. Triumph aimed to gain strategic insights from community-generated content to enhance their visual brand strategy across digital platforms.

Poljects’ Solution

Using advanced AI-powered image recognition, we analyzed over 10,000 Instagram posts tagged with #triumphmotorcycles. Each image was processed to extract visual features—ranging from terrain and lighting to rider posture and background content.

We then used dimensionality reduction and unsupervised clustering to build a multi-dimensional visual landscape, revealing organically emerging content clusters like:

• Clean Bike Shots
• Racing & Action Shots
• Artwork & Conceptual Photography
• On-the-Road Stories
• MC Gear & Accessories

Built using:

• Computer Vision
• Deep Learning
• Dimensionality Reduction
• Clustering Algorithms

Outcome

• Data-driven insights into what drives engagement across Triumph’s social audience
• Optimized content strategy across campaigns and geographies
• AI segmentation now integrated into Triumph’s social listening toolkit

Each cluster was analyzed for engagement uplift—some reaching up to 7,745% higher than average.

Want to transform visual data into strategic insights?

Let’s build a visual segmentation solution for your brand.

Open Innovation Design Collaboration with BMW Motorrad

Key Challenges

• Reignite the interest of a younger, tech-native audience.
• Develop design languages beyond the traditional “macho” aesthetic.
• Integrate connected services and digital experiences into the motorcycle ecosystem.
• Shift from pure mobility to emotional, lifestyle-centered vehicle concepts..

Our Contribution

We supported BMW in the conception, communication, and evaluation of a large-scale innovation contest. The outcome: over 1,500 international submissions—crowdsourced ideas ranging from futuristic electric concepts to fully connected rider ecosystems.

Our design team contributed a set of visionary design proposals based on insights from the contest, focusing on:

• Electric & Autonomous-ready vehicle concepts
• Human-centric, ergonomic design for urban agility
• Smart safety integration
• Adaptive customization & AR-based HUDs
• VR-compatible riding experiences

Visual Highlights

The showcased concepts reflect a radical break from traditional motorcycle archetypes, inspired by both form and function. Merged with cinematic CGI and 3D rendering techniques, these visions became tangible assets for future product development and storytelling.

Impact

• Over 1,000 qualified submissions and 260 consistently active participants
• Global reach across more than 10 countries
• Strong alignment between participant ideas and BMW’s strategic future goals
• Several design concepts now influencing internal product development pipelines

Want to make sense of your publication archive or scientific dataset?

Let’s build an intelligent visual explorer tailored to your domain.

Visualizing 10,000+ Publications on Breast Cancer Diagnosis & Screening

Challenge

This project leverages Natural Language Processing (NLP) and AI-driven clustering to turn unstructured scientific literature into a visually explorable knowledge map. By processing thousands of medical publications related to breast cancer screening, the system uncovers hidden patterns and topic relationships.

Poljects’ Solution

In this interactive dashboard:

• Over 10,000 publications are automatically clustered by semantic similarity
• Significant variables like publication date, title length, and keywords are used to refine exploration
• Topics like “Deep Learning Classification”, “Radiology”, “Covid-19 Pandemic”, and “Racial Interventions” emerge organically
• Users can filter and drill down into clusters to examine publication-level metadata

Built using:

• AI/NLP models for topic modeling & clustering
• Dimensionality reduction techniques (e.g. UMAP, t-SNE)
• Interactive front-end using Plotly + Dash

Outcome

This tool enables researchers, policymakers, and clinical experts to stay ahead of innovation trends, discover knowledge gaps, and accelerate evidence-based decision-making.

Want to make sense of your publication archive or scientific dataset?

Let’s build an intelligent visual explorer tailored to your domain.