Whether you're a seasoned data scientist, a curious beginner or someone eager to understand the impact of data-driven technologies, this event is tailored for you. Build connections with professionals, experts, and fellow enthusiasts, fostering collaboration and knowledge exchange.
Data Science Retreat presents 7 Machine Learning prototypes and projects by Batch 46 participants. The event is free to attend and thanks to SAP Berlin for hosting us.
Agenda:
17:30 - Doors open, Drinks and Networking
18:00 - Welcome and Intro to SAP and SAP LeanIX
18:20 - Introduction by DSR
18:30 - Project Presentation (4 projects)
19:45 - Short break
20:00 - Project Presentation continues (3 projects)
21:00 - Open for networking
21:30 - Wrap up.
PROJECT IDEAS:
1. AirLiness: Proactive Indoor Air Quality Management and Automated Intervention
Project by: Jose Carlos Garay Fernandez
AirLiness is an intelligent indoor air quality management system that uses machine learning to optimize ventilation and air purification. It analyzes real-time and historical indoor sensor data, outdoor environmental conditions, energy prices, and occupant health information to determine the best intervention strategy. The system proactively maintains healthy indoor air while reducing energy consumption and operating costs. By balancing health, comfort, and efficiency, AirLiness supports smarter and more sustainable building management.
2. The RegI (RegulatoryIntelligence)
Project by: Ting Xiang and Anagha Kannampilly Janardhanan
The RegI (RegulatoryIntelligence) is developing a machine learning system designed to convert complex regulatory and governmental information into clear, actionable steps for businesses. This platform acts as a Communication Intelligence Infrastructure, addressing the challenges of fragmented data and highly specialized language that often force companies to hire external consultants. By transforming dense documentation into practical steps, RegI enables organizations to navigate compliance demands more efficiently.
3. AI-Driven Seamless Pattern Generation
Project By: David Terence Watts
This project aims to automate seamless pattern generation for textiles, design, and 3D modeling by eliminating the need for manual editing of repeating image tiles. The system will analyze pattern symmetry and intelligently fill overlapping regions while preserving the original design. It leverages CNNs for image preprocessing, Graph Neural Networks to model symmetric toroidal geometry, and Transformer-based encoders to understand image content. The result is a faster, more accurate, and scalable solution for creating high-quality seamless repeating patterns.
4. Stamp Scout: AI-Powered Stamp Value Detection
Project by: Robin Beck
Stamp Scout is an AI-powered mobile application that rapidly identifies potentially valuable stamps from large collections using computer vision. Instead of searching through hundreds of thousands of catalog entries, it focuses on recognizing visual indicators of rarity and value. The system uses a fine-tuned CLIP vision model to classify stamps based on features such as age, condition, overprints, perforations, and text. It highlights stamps with color-coded value predictions, enabling users to quickly identify items worth further appraisal while filtering out common low-value stamps.
5. Phoneme-Level Mispronunciation Detection for Language Learners
Project by: Ahsan Ali Siddiqui
Pronounc AI is an AI-powered pronunciation coaching system that detects and localizes mispronunciations at the phoneme level. It analyzes a learner's speech recording against the target text using forced alignment and supervised machine learning to identify pronunciation errors. An intelligent feedback module then provides clear, prioritized correction suggestions in plain language. The system delivers precise, actionable guidance, helping language learners improve pronunciation more effectively than traditional score-based tools.
6. Property Lifecycle Simulator for German Rental Investments
Project by: Lina Pavasaryte
This project develops a self-service web application that helps users evaluate German rental property investments using 30-year Monte Carlo simulations. It compares different mortgage fixed-rate options while accounting for taxes, refinancing penalties, recurring costs, and legal regulations. An OCR module automatically extracts data from bank offers and tax documents, while an LLM explains simulation results in simple language. The tool enables informed, data-driven investment decisions without requiring consultation from financial advisors.
7. Context-Conditioned Audio Synthesis for Traditional Iranian Sport
Project by: Mahyar Moghimi
This project develops an AI-powered music generation system tailored for traditional Pahlevani sport training. It fine-tunes an open-source music generation model on a curated dataset of traditional Iranian music to preserve authentic rhythms and musical scales. Users can customize generated tracks by selecting musical styles or traditional poems, ensuring the tempo matches workout requirements. The system enhances training accessibility, personalization, and scalability for both native and non-Farsi-speaking users.
Mark your calendars for July 21st and get ready to be inspired and educated in the realms of Data Science and Artificial Intelligence. Looking forward to having you all at the event.
