Hack4Good begins with a kick-off event at the ETH Student Project House. At the kick-off, the participating non-profits will present selected challenges and the ensuing projects they host with Hack4Good. The participants then form a team around each project. You will also get to meet your teammates and make some first exploratory steps to understand your challenge and the data available.
Over the entire course of the project, you will have a direct contact person on the side of the non-profits whom you can ask questions on the relevant data and discuss ideas with. Your team will also be supported with inputs, tools and feedback by a mentor who is an industry-experienced data scientist. You will meet your organisation contact and mentor at the kick-off as well.
During eight weeks, you and your team mates will organise your project, distribute tasks and manage the project schedule. You will be able to use the facilities of the ETH Student Project House as meeting areas and for work. Hackdays and hacknights occur several times during the course of the programme. All teams come together to work on their projects and there will be plenty of pizza as well as time to exchange your thoughs with other participants.
Your Hack4Good journey is supported by a range of workshops, first of which being the Agile Workshop. You will be exposed to a project management approach that is used for software development. It assists your team to respond to the unpredictability of constructing software. This event will get you started with organising the full project.
Here, you will learn the theoretical concepts behind impact evaluation and apply them directly on your project. Some questions you will have to answer: How can the impact of your project be defined? Which indicators can be used to measure the impact of your work?
The title says it all: you will learn about what an effective pitch consists of – both in theory and practice.
In the end of our programme the teamswill present their work.
You will learn about theobjectives of the projects, the challenges faced on the way, and, of course,how the teams leveraged their data science skills to support non-profitorganisations and create real-world impact.
After eight weeks, you and your team will hand over your product to the non-profit organisation. The developed code will be open-source and uploaded to a git repository together with your documentation. Each of the participants will receive a certificate of participation. Finally, you get to celebrate your achievement with your fellow participants and all engaged parties in a social event. Get to know each other and have fun!
Build an AI-powered platform that allows non-technical Caritas staff to convert raw tabular data (CSV, Ex-cel, SQL) into clear, interactive dashboards by uploading data and asking questions in natural language.The system should detect appropriate chart types, generate numeric summaries and descriptive insights,and assemble multiple charts into cohesive dashboards. The platform must support English, German, andFrench. Open-source LLMs are preferred to ensure cost efficiency and data privacy.
Key skills: RAG, open-source LLMs, data visualisation, multilingual NLP, Python
Design automated controls to detect procurement non-compliance and fraud risk patterns within HI’s lo-gistics information system. The solution should translate procurement rules into machine-readable checks(threshold breaches, required quotations, date consistency), identify anomalies such as artificial splitting ofpurchases, supplier concentration, and document inconsistencies, and produce alerts classified by severitywith traceable justification. The proof-of-concept must be tested on real procurement datasets.
Key skills: Rule-based systems, anomaly detection, text similarity, NLP, compliance analytics
Evaluate different unsupervised machine learning and deep learning approaches to automate the detec-tion of potentially fraudulent behaviours during humanitarian field data collection. Building on an initialHack4Good 2022 project, this edition leverages significantly larger and more standardised datasets. IM-PACT provides clean, well-structured data so the team can focus on feature engineering, modelling, andevaluation rather than data cleaning.
Key skills: Unsupervised ML, deep learning, anomaly detection, feature engineering
The UN Energy Compacts registry contains 160+ voluntary commitments in heterogeneous formats(Word/PDF). The goal is to develop a standardised, machine-readable dataset by migrating all Compactsinto a harmonised Google Sheets template, integrate annual progress survey data, and build an internalmonitoring dashboard. The system should allow automated or semi-automated annual updates and sup-port country-level and sector-level tracking.
Key skills: Data engineering, data harmonisation, document parsing, dashboard development
Monitor the ratio of animal-based vs. plant-based proteins in the online shops of Migros and Coop, Switzer-land’s two largest retailers (80% market share). The project involves web scraping product data (name, price,promotion status, food group classification), calculating the protein split using the WWF method adaptedby Greenpeace, and developing a repeatable weekly monitoring procedure. Current data shows a 90/10animal-to-plant ratio, far from the 40/60 target needed for net-zero by 2050. Results will inform advocacy,public communication, and retailer engagement.
Key skills: Web scraping, data analysis, automation, Python
A combined project tackling two challenges for AsyLex’s Rights in Exile (RiE) platform, which serves 3,800+refugees and legal advisors annually. (1) UI/UX Redesign & Data Management: Streamline the Word-Press/Elementor website with improved navigation, mobile-first design, and performance optimisation; es-tablish a single source of truth for 1,500+ database entries by replacing manual duplication with automatedsync pipelines. (2) AI Legal Chatbot: Build an AI-powered legal assistant that helps asylum seekers andlawyers find relevant information, answers questions based on platform content, and directs users to theright resources across multiple languages.
Key skills: UI/UX, WordPress, data pipelines, chatbot development, NLP, RAG
Develop a prototype AI-assisted tool that helps users evaluate whether projects contribute to the circulareconomy according to BASE Foundation’s categorisation framework, used across Latin American coun-tries. The tool should analyse project descriptions using NLP, identify circular economy strategies (reuse,repair, recycling, industrial symbiosis), apply structured evaluation filters (inclusion, exclusion, safeguards,circularity level), and generate transparent evaluation summaries. The prototype should highlight missinginformation and be designed for non-technical users.
Key skills: NLP, document analysis, classification, LLM prompting, Python
Build an AI chatbot that guides IAMANEH’s partner organisations through annual reporting interactively.Staff upload project documents to generate tailored questionnaires; partners answer conversationally viaa shareable link with text and voice input in English, French, and German. The chatbot evaluates answerquality in real time, requesting clarification when responses are off-topic, superficial, or unclear. Sessionpersistence allows partners to pause and resume. The tool aims to eliminate the time-consuming back-and-forth that currently delays reporting cycles.
Key skills: Conversational AI, NLP, speech-to-text, LLM orchestration, multilingual
Build an automated pipeline that, for a given science/technology/innovation policy theme, discoversand links relevant academic publications (via OpenAlex, 250M+ works) and EU institutional documents(via data.europa.eu), then generates structured, policy-relevant insights using semantic search and LLMprompting. Where publications explicitly reference specific policy initiatives tracked in STIP Compass, thepipeline should detect and capture these direct links. Deliverable: a working Python prototype or Streamlitweb app.
Key skills: API engineering, semantic search, embedding models, LLM prompting, Python
