Data Science Retreat Demo Day #47

Překládám do češtiny…

**Hey Berlin Data Folks!** We’re looking forward to welcoming you to our next **Data Science Demo Day on 20th October**. Our **Batch 47 participants** have been busy working on their final projects, and they’re ready to share what they’ve built. This time, we have **7 projects** covering a pretty wide mix of topics, from groundwater exploration and healthcare to career transitions, engineering, agriculture, fire detection, and even fermentation. It’ll be a relaxed autumn evening to see some practical data science and AI applications, meet others from Berlin’s data community, and chat about ideas, projects, and what people are working on. And, of course, we’ll have some pizza and drinks to keep everyone going. **Free to Attend!** **Agenda:** **17:30** \- Drinks and Networking **18:00** \- Welcome & Introduction Followed by Project Presentations **Project Ideas:** **1\. DeepWell \- Machine learning for predictive groundwater mapping** ***Project by Sascha Alexeyenko*** Finding reliable groundwater is a major challenge in regions where water is scarce, and drilling a borehole can be expensive without any guarantee of success. DeepWell explores how machine learning can help identify promising locations for groundwater drilling. By combining historical borehole records, including unsuccessful attempts, with geological, terrain, and climate data, the project aims to predict the likelihood of finding water with sufficient yield and quality. The initial focus is on Kenya, particularly Turkana and Marsabit counties, where access to reliable water remains a challenge. The goal is to help make drilling decisions more informed, reduce unsuccessful attempts, and support better groundwater planning. **2\. Personalized Psychiatric Treatment Selection for Depression** ***Project by Aleksandra Garifulina*** Finding the right antidepressant treatment can be a very individual process. This project explores whether machine learning can help support more personalized treatment selection using patient-level clinical characteristics and treatment-response data. The project looks at supervised classification methods to match treatments more closely to individual patient profiles. It also explores how clinical guidelines, treatment suitability, and contraindications could be incorporated into the process. The goal is to investigate how these tools could support clinical decision-making rather than replace it. **3\. Job Transition Workspace** ***Project by Timothy LeBon*** Changing careers can mean juggling job searches, applications, skills, resumes, advice, and a lot of uncertainty all at once. This project brings those pieces together in one intelligent career transition workspace. Using data from O\*NET, the U.S. Current Population Survey, and the Anthropic Economic Index, the system recommends roles based on a person’s skills, experience, career goals, and job-market demand. It also includes an LLM workspace that can work with resumes, chats, and cover letters, along with tools for things like job searches, email, reminders, and calendar management. The goal is to create one place where people can plan and manage a career change with more personalized, data-informed support. **4\. AI\-Assisted Requirement Change Impact Analysis** ***Project by Lakshmi Swetha Pasupuleti*** In complex engineering projects, changing one requirement can affect many other things, from software requirements and design decisions to test cases. This project looks at how AI can help engineers understand those connections, especially when they are not already documented. The idea is to use semantic analysis to find related engineering artifacts and, importantly, explain why they might be affected by a change. The project will compare AI-based methods with more traditional traceability approaches across different change scenarios. The end goal is a lightweight prototype that explores where AI can make requirement change analysis easier and more useful for engineers. **5\. cropSpectAI: Spectral Sensing and AI\-Driven Crop Health Monitoring** ***Project by Sharana Shivanand*** What if farmers could spot changes in a plant’s health before those changes were visible to the eye? cropSpectAI explores this idea using a low-cost portable device that combines spectral sensing with machine learning. The system captures how leaves reflect light across different wavelengths and uses those patterns to estimate important plant traits such as chlorophyll and nitrogen content. The project experiments with several machine-learning approaches to see how affordable, edge-based technology could help farmers identify problems earlier, make better-informed decisions, and potentially reduce unnecessary use of resources. **6\. Residential Fire and Smoke Detection Platform** ***Project by Abhishek Acharya*** Traditional smoke and heat detectors are essential, but they usually react once smoke or heat has reached the sensor. This project explores whether computer vision can provide an additional layer of information by recognizing signs of fire and smoke visually. The system is designed to understand context too. It looks at situations involving cooking flames, candles, fireplaces, steam, cigarette smoke, pans, stoves, and people to help distinguish an actual hazard from normal household activity. The aim is not to replace certified smoke or heat detectors, but to explore how visual AI could complement them with earlier, more context-aware detection while reducing unnecessary alarms. **7\. Predictive Fermentation Assistant** ***Project by Sylas Lau*** Fermentation sounds simple: give microbes the right ingredients and conditions, then give them time. In reality, factors like temperature, moisture, oxygen, and timing all interact, which makes getting consistent results surprisingly tricky. This project explores how machine learning can help home fermentation enthusiasts better understand and manage that process. From bread and yoghurt to koji, tempeh, and kimchi, the idea is to use data to make fermentation a little more predictable while still keeping the hands-on nature that makes it interesting. **20:10** \- Open for Networking **20:30** \- Wrap Up Come by, grab some pizza and a drink, meet the people behind the projects, and spend a cosy October evening with Berlin’s data community. We have **limited seats**, so please RSVP soon.