ExoExplorer

Inspiration

Inspired by the HackTX 2025 theme “Build Beyond the Stars, One Story at a Time,” we wanted to create something that captures the wonder of discovery while staying grounded in real science. ExoExplorer was our way of transforming complex exoplanet research into an experience anyone could enjoy, explore, and learn from. We envisioned a tool that lets people see the universe not as distant data points, but as living stories of potential worlds — each with its own unique environment, orbit, and possibility for life. By combining the power of AI with NASA’s exoplanet data, we built a platform that invites both scientists and everyday dreamers to go beyond the stars and uncover their own story among them.

What It Does

ExoExplorer is an interactive web application that lets users:

Search and explore confirmed and candidate exoplanets using data from NASA’s Kepler and KOI (Kepler Objects of Interest) catalogs.
Visualize habitability through dynamic graphs showing temperature, pressure, orbital speed, and distance from Earth.
Compare each planet to Earth with radar plots and interactive statistics that make astrophysical concepts intuitive.
Ask questions conversationally via Gemini 2.5 Flash, which acts as an AI guide explaining each planet’s habitability and features in plain language.
Generate realistic NASA-style renders of planets using Gemini’s multimodal image capabilities to produce scientifically accurate, labeled planetary renders with soft lighting and detailed textures.
Use Agentic AI planning that outlines how an AI would retrieve, analyze, and visualize exoplanetary data in a multi-step reasoning process.

All in all, ExoExplorer turns data into discovery: a platform where science, computation, and art meet.

How We Built It

Frontend: Built with TypeScript, JavaScript, React, and TailwindCSS (shadcn/ui) for a clean, responsive, and accessible user interface.

Backend: Developed using Flask and integrated with Gemini 2.5 Flash (via the @google/generative-ai library) for both text-based explanations and image generation. Flask handles API endpoints that process requests, interact with the Gemini model, and serve data securely to the frontend.

Data: Local Kepler and KOI (Kepler Objects of Interest) CSV datasets were parsed and normalized using Python and Node-based utilities for consistency and fast lookup.

Visualization: Interactive data visualizations were built using Recharts and Framer Motion to create smooth, dynamic charts and planet comparisons.

Data Processing: We implemented a lightweight habitability scoring heuristic defined as H=100×(0.35R+0.5T+0.15P) where 𝑅 measures Earth-radius similarity, 𝑇 represents temperature proximity to 288 K, and 𝑃 represents atmospheric pressure near 1 bar.

AI Integration: All AI generation, from natural-language explanations to NASA-style planet renders, is handled safely on the server through Gemini 2.5 Flash, ensuring secure API key management and efficient multimodal responses.

Every user action, including searching, comparing, asking questions, and generating images, flows through a modular system that unites astrophysical data, AI reasoning, and visualization in one cohesive experience.

Challenges We Ran Into

Data normalization: NASA’s Kepler and KOI datasets contain thousands of columns with inconsistent naming conventions and missing values. Cleaning and mapping these into a usable JSON schema was one of our first major challenges.

Scientific visualization: Creating UI components that balanced accuracy with accessibility was difficult, especially when making logarithmic temperature scales readable to non-scientists.

Image generation fidelity: Getting Gemini 2.5 Flash to consistently follow physically based rendering (PBR) constraints and NASA-style composition required careful prompt design and iterative refinement.

API security: We had to ensure that Gemini API keys remained secure while still supporting real-time AI interactions from the browser.

Performance: Parsing and serving large CSV datasets from the backend while maintaining fast, responsive frontend performance demanded optimization and caching strategies.

Accomplishments

Created a fully interactive AI-driven exoplanet explorer that allows anyone to explore, compare, and visualize potentially habitable worlds regardless of scientific background
Integrated Gemini 2.5 Flash for both text-based explanations and image generation, showcasing a dual-modality AI use case in planetary science
Built a complete agentic pipeline (Retrieve → Analyze → Explain → Visualize) that emulates how an AI researcher would study and present exoplanetary data
Designed and refined NASA-grade rendering prompts that generate scientifically accurate, hyperrealistic single-planet visuals aligned with NASA’s aesthetic and lighting standards
Developed a data-rich, visually intuitive interface that makes complex astrophysical information understandable and engaging for both scientists and the general public

What We Learned

Learned that effective prompt design requires both creativity and scientific precision, as physics-informed phrasing produced more accurate and realistic Gemini outputs
Discovered the importance of accessibility in science communication, improving our ability to design interfaces that make complex astrophysical data understandable to everyone
Realized how AI can democratize scientific discovery by transforming open datasets into intuitive, human-centered insights about the universe
Found that using a single multimodal model for both text and imagery enables new forms of scientific storytelling and data visualization

What's Next for ExoExplorer

Expand dataset coverage to include JWST and TESS exoplanet discoveries, integrating live API syncing for continuously updated planetary data
Implement real-time Gemini chat streaming to enable continuous, conversational exploration as users interact with different exoplanets
Add 3D WebGL visualizations featuring orbit paths, rotational dynamics, and light-curve simulations for more immersive scientific experiences
Develop an “AI Observatory Mode” where Gemini can simulate atmospheric composition, potential biosignatures, and stellar flux variations in real time
Introduce user personalization options that allow individuals to bookmark planets, track discoveries, and generate custom PDF reports
Collaborate with NASA and other international space agencies to integrate verified astrophysical data, enabling ExoExplorer to evolve into a more advanced, research-grade platform capable of generating scientifically tangible insights and predictive models