Data Science capstone projects batch #32
by Ekaterina Butyugina
We’re thrilled to celebrate the achievements of our latest graduates from Zurich’s Batch #32 and Munich’s 11th cohort, who have just wrapped up their Data Science journey with two remarkable, real-world projects.
This round of final presentations showcased how data science and AI can drive tangible impact across industries, from transforming business development workflows to reinventing the way educators design learning experiences.
Take a look at how our graduates are using data science to generate insights, push boundaries, and create real-world impact.
BAK Economics is a Swiss economic research institute that produces comprehensive economic studies and makes the resulting findings available to politics, business and society. Their challenge was a highly manual and inefficient process: searching over 10 platforms like Simap and EU Tender, spending hours daily (specifically, 3 hours per day) to sift through more than 850 tenders each month. With over 90% of these tenders being irrelevant, the team faced wasted time, missed opportunities, and even lost 6-figure deals.
The solution is an Intelligent Tender Matching System. The system's goal is to automate the entire discovery process, filter out the irrelevant "noise," and present a curated list of high-potential opportunities directly to the BAK team.
The approach involves three main steps. First, automated data collection uses APIs and scrapers to pull data from Swiss (SIMAP), German (e-VERGABE), and European (EU tender), building a standardized database of historical tenders. Second, an AI-powered analysis engine, using GPT-4o-mini, analyzes each tender. This AI was trained on BAK's specific needs, including a 4,100-word manual, their service portfolio, and 93 examples of winning tenders. Finally, the system delivers curated recommendations.
The result is a highly efficient system that transforms the team's workflow. For example, the automated process analyzes 850 tenders and identifies 20-30 relevant opportunities ready for review. This system achieved a 95% accuracy rate and successfully detected missed opportunities from the past. See the picture below.
The impact of this system is substantial. It reduces manual effort from approximately two hours per day to a concise 15-minute review, ensuring comprehensive coverage across all tender platforms. By minimizing missed opportunities and eliminating repetitive administrative tasks, it enables BAK Economics’ experts to redirect their focus from manual searches to high-value strategic decision-making.
Students: Karlo Lukic, Debbie To
Constructor Tech is an education technology company focused on advancing the use of artificial intelligence in learning design and delivery. They aim to address one of the most persistent challenges in education: the time-intensive and inconsistent process of creating comprehensive, high-quality course content. Traditional AI-assisted tools, while fast, often generate superficial or unstructured materials that fail to meet pedagogical standards. Educators spend countless hours refining outlines, ensuring logical progression, and maintaining instructional coherence, a process that limits scalability and innovation in course development.
To tackle this challenge, Karlo and Debbie developed Coursely, a multi-agent AI system designed to redefine how educators create and structure high-quality educational content. By orchestrating multiple specialized AI agents that collaborate on course planning, content generation, and quality review, Coursely introduces a systematic, scalable, and pedagogically sound approach to course design.
As the team noticed, both humans and AI share a fundamental limitation: they can't effectively focus on many complex aspects at once. Even with extensive prompting, a single AI agent struggles to maintain quality across many dimensions of course design.
Coursely's breakthrough lies in its multi-agent architecture, inspired by how human teams collaborate on complex projects. Rather than relying on a single AI agent, the system employs specialized AI agents working in concert:
The system leverages the Claude Agent SDK library for Python to create a well-organized team of agents. Key features include:
As future possibilities, the team envisions exciting enhancements of the Coursely multi-agent system:
Coursely represents more than just workflow automation. It’s a paradigm shift in educational content creation! By combining AI with organizational principles of human collaboration, the so-called human-in-the-loop, it empowers educators to focus on what matters most: inspiring and guiding their students.
As this technology evolves, it promises to democratize access to high-quality educational content creation, making it possible for educators to develop comprehensive, engaging courses that truly serve their students' needs.
A huge thank you goes to our partners, mentors, and instructors who guided the teams throughout their journeys. Your insights and collaboration helped turn ideas into working prototypes with real-world value.
To our graduates, your persistence, curiosity, and drive to solve meaningful problems are what make this community so special. We can’t wait to see where you’ll go next and how you’ll continue shaping the future of AI and data-driven innovation.
If you’re inspired by these stories, discover how you can join our next Data Science cohort at Constructor Nexademy and start building your own impact-driven project.
This round of final presentations showcased how data science and AI can drive tangible impact across industries, from transforming business development workflows to reinventing the way educators design learning experiences.
Take a look at how our graduates are using data science to generate insights, push boundaries, and create real-world impact.
Finding the Needle in the Haystack: Automating Tender Discovery for BAK Economics
Students: Damla Ünal, Ana Dobson, Cedric Jansen, Tim HasenfussBAK Economics is a Swiss economic research institute that produces comprehensive economic studies and makes the resulting findings available to politics, business and society. Their challenge was a highly manual and inefficient process: searching over 10 platforms like Simap and EU Tender, spending hours daily (specifically, 3 hours per day) to sift through more than 850 tenders each month. With over 90% of these tenders being irrelevant, the team faced wasted time, missed opportunities, and even lost 6-figure deals.
The solution is an Intelligent Tender Matching System. The system's goal is to automate the entire discovery process, filter out the irrelevant "noise," and present a curated list of high-potential opportunities directly to the BAK team.
The approach involves three main steps. First, automated data collection uses APIs and scrapers to pull data from Swiss (SIMAP), German (e-VERGABE), and European (EU tender), building a standardized database of historical tenders. Second, an AI-powered analysis engine, using GPT-4o-mini, analyzes each tender. This AI was trained on BAK's specific needs, including a 4,100-word manual, their service portfolio, and 93 examples of winning tenders. Finally, the system delivers curated recommendations.
The result is a highly efficient system that transforms the team's workflow. For example, the automated process analyzes 850 tenders and identifies 20-30 relevant opportunities ready for review. This system achieved a 95% accuracy rate and successfully detected missed opportunities from the past. See the picture below.
The impact of this system is substantial. It reduces manual effort from approximately two hours per day to a concise 15-minute review, ensuring comprehensive coverage across all tender platforms. By minimizing missed opportunities and eliminating repetitive administrative tasks, it enables BAK Economics’ experts to redirect their focus from manual searches to high-value strategic decision-making.
Coursely: Multi-Agent Magic for Educational Content
Students: Karlo Lukic, Debbie ToConstructor Tech is an education technology company focused on advancing the use of artificial intelligence in learning design and delivery. They aim to address one of the most persistent challenges in education: the time-intensive and inconsistent process of creating comprehensive, high-quality course content. Traditional AI-assisted tools, while fast, often generate superficial or unstructured materials that fail to meet pedagogical standards. Educators spend countless hours refining outlines, ensuring logical progression, and maintaining instructional coherence, a process that limits scalability and innovation in course development.
To tackle this challenge, Karlo and Debbie developed Coursely, a multi-agent AI system designed to redefine how educators create and structure high-quality educational content. By orchestrating multiple specialized AI agents that collaborate on course planning, content generation, and quality review, Coursely introduces a systematic, scalable, and pedagogically sound approach to course design.
As the team noticed, both humans and AI share a fundamental limitation: they can't effectively focus on many complex aspects at once. Even with extensive prompting, a single AI agent struggles to maintain quality across many dimensions of course design.
Coursely's breakthrough lies in its multi-agent architecture, inspired by how human teams collaborate on complex projects. Rather than relying on a single AI agent, the system employs specialized AI agents working in concert:
- Main Agent: Orchestrates the entire process of course design
- Planner: Develops comprehensive plans before execution
- Source Processor: Handles any background materials for a course
- Content Generator: Produces detailed lessons
- Five Specialized Reviewers: Each focusing on critical aspects (outline coherence, inclusivity, learning paths, instructional quality, structural integrity) of course design
- Course Editor: Implements feedback from specialized reviewers and ensures consistency in course design
The system leverages the Claude Agent SDK library for Python to create a well-organized team of agents. Key features include:
- Planning Mode: Develops comprehensive plans before execution
- Automated Iteration Loops: Reviewers provide feedback that triggers improvements
- Focused Expertise: Each agent concentrates on its specialized domain
- Seamless Coordination: The Claude Agent SDK handles state management and message routing
As future possibilities, the team envisions exciting enhancements of the Coursely multi-agent system:
- Extended customization through Model Context Protocol (MCP) servers
- LMS integration for seamless deployment at educational institutions
- Expanded content types: Interactive quizzes, video scripts, project assignments
- Adaptive learning paths based on student needs
Coursely represents more than just workflow automation. It’s a paradigm shift in educational content creation! By combining AI with organizational principles of human collaboration, the so-called human-in-the-loop, it empowers educators to focus on what matters most: inspiring and guiding their students.
As this technology evolves, it promises to democratize access to high-quality educational content creation, making it possible for educators to develop comprehensive, engaging courses that truly serve their students' needs.
Conclusion
As we close the chapter on the Data Science Final Projects for Batch #32, we couldn’t be prouder of what our students accomplished. These projects prove that when technical skill meets imagination, data science becomes a powerful force for innovation.A huge thank you goes to our partners, mentors, and instructors who guided the teams throughout their journeys. Your insights and collaboration helped turn ideas into working prototypes with real-world value.
To our graduates, your persistence, curiosity, and drive to solve meaningful problems are what make this community so special. We can’t wait to see where you’ll go next and how you’ll continue shaping the future of AI and data-driven innovation.
If you’re inspired by these stories, discover how you can join our next Data Science cohort at Constructor Nexademy and start building your own impact-driven project.
