Inspiration
In 2025, I was sitting in my room in Senegal, asking myself: how can I become the person I want to be, and how can I help my community? I love astronomy deeply, but I struggled to find accessible resources that could help me truly understand it. So I asked myself: why not build something that helps me — and everyone facing the same problem? That moment gave birth to Galicea. The name itself is personal: I created it by combining Alice (one of my closest friends) and Galaxie — Alice + Galaxie = Galicea.
What Galicea Does
Galicea is an AI-powered astronomy assistant that can:
- Greet users and remember their name and location
- Calculate distances between the user and celestial objects
- Describe planets and define astronomical terms
- Identify which artificial satellites are currently visible from the user's location in real time
- Correct errors in user questions
- Honestly say when it doesn't know something — it never invents answers
What I Learned
Building Galicea taught me Retrieval-Augmented Generation (RAG) using LangChain, vector storage with ChromaDB, and real-time astronomical calculations using Astropy and the Gaia DR3 database. Star distances are calculated using the parallax formula:
$$d = \frac{1}{\pi}$$
where $d$ is the distance in parsecs and $\pi$ is the parallax in arcseconds.
How I Built It
Before writing a single line of code, I opened my notebook and wrote down everything I wanted Galicea to be and do. Then I organized my schedule, chose my tools, and started building step by step.
Tech stack:
- Python — core language
- LangChain — RAG pipeline
- ChromaDB — vector database (14,000+ indexed chunks from scientific PDFs)
- Groq API + LLaMA 3.3 — language generation
- HuggingFace Embeddings — semantic search
- Astropy + Geopy — real-time celestial calculations
- Streamlit — web interface
Challenges
The journey was far from easy. I faced package conflicts, mysterious errors, and wrong code lines that took hours to debug. At one point, my computer ran out of storage completely — I could no longer run or even open my program. It was devastating. I stopped coding for months. I was desperate. But I came back. I solved the storage problem by migrating from a local model (Ollama/Mistral) to the Groq API, moving computation to the cloud. My laptop stopped overheating. Galicea came back to life. I built this entirely alone, with no mentor, no formal training — just documentation, trial and error, and a deep desire to build something meaningful for my community.
Accomplishments that we're proud of
- Built a fully functional AI astronomy assistant from scratch, alone, at 17 years old in Senegal with no formal training
- Indexed 14,000+ chunks from scientific PDFs to create Galicea's knowledge base
- Integrated real-time celestial calculations using Astropy and the Gaia DR3 database
- Selected as Finalist in the AI Futures Challenge 2026 by Northeastern University London, awarded a £1,000 scholarship
- Successfully migrated from a local LLM to cloud-based Groq API, solving overheating and storage issues
What we learned
- How to build a RAG pipeline from scratch using LangChain and ChromaDB
- How to work with real astronomical data from the Gaia DR3 database
- How to secure API keys and manage environment variables properly
- How to structure a complex Python project independently
- That persistence matters more than perfection — I stopped for months and came back stronger
What's next for Galicea AI
- Adding real-time tracking of artificial satellites using TLE data from CelesTrak and the Skyfield library
- Expanding the knowledge base with more scientific papers and multilingual support for French and Wolof speakers
- Deploying Galicea online so students across Africa can access it freely
- Adding a feature to generate personalized astronomy learning paths based on the user's level and interests
- Submitting Galicea to more competitions and research programs to grow its impact
Log in or sign up for Devpost to join the conversation.