AKSA: What if AI got us to look up at the sky instead of down at our screens? 🌌

Yash Chandra — Mon, 12 Oct 2026 06:58:17 +0000

Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass Submission 🌿

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

AKSA is an AI-powered stargazing companion designed to turn curiosity about the night sky into a reason to go outside.

Have you ever looked up at the night sky and wondered what you were seeing? Is that a planet? Where should you look to find Saturn? Is tonight a good time to stargaze?

That's the experience I wanted to make more accessible.

Most AI experiences encourage us to spend more time looking at screens. With AKSA, I wanted to explore the opposite idea: what if AI helped us put our phones down and discover the real world?

AKSA combines astronomy calculations with a mobile interface and local AI, helping people explore celestial objects, understand what's visible, and learn about the universe.

✨ What can you do with AKSA?

🌙 Tonight Verdict: Find out whether conditions are suitable for stargazing at your location.

🔭 Find Objects: Explore celestial objects with directions, altitude, rise and set times, and star-hopping hints.

🧭 Sky Detective: Explore an interactive sky map and use clues to identify objects.

🌠 Stargazing Missions: Follow short observing challenges that encourage you to step outside and look up.

🧠 On-Device AI: Use a local language model to turn computed astronomical information into more conversational explanations.

The goal isn't to replace the night sky with an AI-generated answer. It's to help people become curious about what they can actually observe.

Demo

🎥 video demo:

picturetolink.com

🚀 Try AKSA: Open the live web demo-

aksa-mobile.onrender.com

📱 Android application: https://github.com/YashChandra141/AKSA-mobile/releases/tag/AI

Code

AKSA-mobile is open source. Explore the code, inspect the implementation, and see how the mobile experience is built.

📱 GitHub repository:

YashChandra141 / AKSA-mobile

AKSA a mobile star gazing companion

🌌 AKSA — Your Stargazing Companion

Know whether tonight is worth going out. Discover what's actually visible in your sky. Find it with directions you can follow using your own hands.

AKSA is an offline-first, AI-powered stargazing companion that turns your phone into a personal guide to the night sky.

Unlike a generic AI astronomy chatbot, AKSA calculates astronomical positions on-device. Its local language model explains the results in friendly, natural language — it never invents celestial positions or decides what is visible from memory.

No account. No telemetry. Astronomy computed on your device.

📸 Screenshots

Tonight Verdict	Sky Detective	Object + AI chat

Missions	Models

✨ Features

🌙 Tonight Verdict

Should you go stargazing tonight?

AKSA calculates a GO / MAYBE / SKIP recommendation using:

Astronomical darkness, based on the Sun being below −18°.
Real cloud-cover data.
Moon phase and how long the Moon will be above the horizon.
…

View on GitHub

The project combines mobile development, astronomy functionality, and experiments with local AI inference.

How I Built It

AKSA-mobile brings together a mobile application, astronomy-related calculations, and on-device AI.

🛠️ Technology stack

React Native + Expo: Mobile application and interface.

TypeScript: Application logic.

Python: Astronomy/backend functionality.

GGUF: Format for compatible open-weight language models.

llama.rn / llama.cpp: Local inference approach for supported devices.

🧠 Separating astronomy from AI

One of the important design decisions is separating astronomical calculations from language generation.

The astronomy engine computes the underlying information, such as celestial positions and observing details. The language model helps explain that information in a conversational way.

This matters because an AI model shouldn't be treated as the authoritative source for astronomical calculations when dedicated software can compute them.

📱 Why local inference?

The app includes a model manager for downloading or importing compatible GGUF models.

I've been experimenting with smaller models that are practical for mobile hardware, including SmolLM2-360M for fast testing and Gemma 3 1B as a larger model option.

Running models locally introduces real engineering challenges: memory limits, inference speed, model compatibility, and the balance between answer quality and performance.

But it also creates opportunities to build AI experiences without requiring a hosted model API for every interaction.

Why Does Open Innovation Matter?

Open innovation gives developers the freedom to experiment beyond a single proprietary AI endpoint.

For AKSA, open-weight models and open-source inference tools make three things possible.

AI closer to the user

Local inference creates a path toward answering questions directly on the device, reducing the need to send every prompt to a remote AI provider.

Freedom to experiment

I can explore different models, runtimes, and performance trade-offs, including how smaller models behave on everyday Android hardware.

AI that encourages real-world discovery

The purpose of AKSA isn't to keep users chatting indefinitely. It's to make astronomy more approachable and encourage people to experience the night sky for themselves.

Open innovation makes it possible to explore this idea while keeping the implementation open for other developers to study, improve, and build upon.

Prize Categories

Primary Challenge:Touch Grass — Open-Source AI Challenge

AKSA-mobile explores open-weight models and local inference in an Android astronomy companion designed to encourage real-world discovery.

Best Use of Render
Render deployment configuration is included for hosting a static site. A verified public deployment URL will be added after end-to-end deployment testing .

devchallenge #hf26challenge #OpenSource #AI #LocalAI #Astronomy #ReactNative #TouchGrass

DEV Community: Yash Chandra