Ofrenda🕯️:Remembering Loved Ones With Gemma

Dezeena Priyadarshy — Sun, 01 Nov 2026 04:02:29 +0000

What I Built?

Ofrenda is a small app that helps write a remembrance for a loved one
for Día de los Muertos. You enter their name, their favorite foods,
a song they loved, and one memory. Gemma then writes a short, warm
message to place on a virtual ofrenda.

I chose this because nothing haunts us more than a loved one's memory. Ghost in the Machine felt like a
chance to build something gentle instead of scary.

Demo

https://dev.to/challenges/hacktoberfest-week4-2026-10-26)*

Code

https://github.com/ketone834-sourcs/ofrenda-gemma]

How I Built It

Model: Gemma 3 (4B), an open-weight model from Google
Runtime: Ollama, serving the model through a local API
Interface: Gradio, for a simple form-and-result page
Environment: a Google Colab notebook with a free T4 GPU
A system prompt tells the model to use only the details given and never to invent facts, names or dates. That matters because a made-up detail about someone's loved one would be hurtful. I built this entirely on an Android phone, using only a browser.

Why Does Open Innovation Matter?

These stories are personal. With an open-weight model, a family's memories don't have to be sent to a company's servers or locked in a service that might disappear someday.

To be transparent: I used Colab for my demo, since my phone doesn't have enough memory to run the model. But because Gemma is open and Ollama is open source, the same code runs fully offline on any computer. A closed API would not allow

The Spirit of It

I tested Ofrenda with a made-up grandmother, Sebati, who loved Rasmalai and hummed "Jalebi baby" while shelling peas on the porch. I typed in a handful of small details, and Gemma gave them back
as something that read like a tender note.I expected a machine to feel like the ghost in this challenge: something cold and haunting. Instead, it felt closer to a candle on an ofrenda, a small light that helps you remember. The people we miss live in details like foods, songs and Sunday habits, and that is exactly.whatthis app asks for.

Prize Categories

<!-- Which partner categories are you entering? List
Best use of Gemma

SlothSum 🦥: Two Sentences, Zero Effort, Powered by Open-Weight Gemma

Dezeena Priyadarshy — Fri, 23 Oct 2026 06:49:54 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I built?

SlothSum is a summarizer for the wonderfully lazy. Paste in a sprawling message, an article or a rambling email, and it hands back exactly two crisp sentences. No scrolling, no skimming, no guilt. It is for anyone who is drowning in long texts and would rather spend that energy on a nap.

Demo
https://youtube.com/shorts/2EAS38pL0FQ

Code
https://github.com/ketone834-source/slothsum

How I Built It?

I built the whole project on an Android phone, with no laptop. In a free Google Colab notebook I installed Ollama, an open-source runtime for language models, and pulled Gemma 3 (1B), Google's open-weight model. A short Python script wraps the pasted text in a prompt that casts the model as a lazy sloth and demands a two-sentence summary. The entire project is under ten lines of code, because the open tooling does the heavy lifting.

****Why does open innovation matter for what you built?

Open weights made this project possible for me. Without a credit card, a computer or an API key, I could still download a capable model and run it myself. There was no account to approve, no per-request bill and no usage cap to watch. Because the model is just a name in one line of code, I can swap Gemma for another model by changing a single word, and I can read and modify everything. For a first-time builder with only a phone, an open ecosystem was the difference between building something and watching from the sidelines.

The Lazy Payoff

It takes me too much time and frustration to both read and understand those long paragraphs especially during needy time like a quick revision before exam.With two line summarizarion it saved my time like just 5 minutes and I could understand it in 5 minutes rather than hours it took when reading those long paragraphs.A 1B-parameter model is small, so its summaries are quick and good enough for a gist, though not flawless.

Prize Categories
Best use of gemma

DEV Community: Dezeena Priyadarshy