Fluently — Your Personal English-Learning Companion

palakshi verma — Wed, 21 Oct 2026 06:15:41 +0000

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

What I Built

Fluently — Your Personal English-Learning Companion

Learning English consistently can be challenging when you need to manage classes, track attendance, and find time for daily practice. Fluently is designed to bring these activities together in one place and make building an English-learning habit easier.

Fluently aims to help learners organise their learning routine, practise English regularly, and get help from an AI teacher. Instead of spending too much time deciding what to practise next, learners can focus on improving their communication skills, one step at a time.

The idea is simple: make consistent English practice easier, so learners can spend less time organising and more time learning.

Demo

🚀 Try Fluently: http://fluently-4lhh.onrender.com/

Code

💻 GitHub Repository: https://github.com/PalakshiVerma/Fluently

How I Built It

Fluently is built around Python and Streamlit, with the goal of integrating an AI teacher powered by an open-weight Gemma model through a hosted inference API.

The project architecture uses the following technologies, where implemented:

Python + Streamlit: Application logic and user interface.
Gemma: AI-powered assistance for English-learning questions and follow-up conversations.
SQLite: Structured data such as classes, attendance, and practice progress.
PyMuPDF and python-docx: Extracting text from learning materials, if the materials feature is included.
ChromaDB and embeddings: An optional retrieval layer for answering questions based on uploaded notes.
Render: Hosting the web application.

The focus is on using AI to make English practice more accessible and easier to maintain, rather than building a generic chatbot without a specific purpose.

Why Does Open Innovation Matter?

Open innovation gives developers the opportunity to experiment, learn, and build specialised solutions using openly available models and tools.

For Fluently, open-weight models such as Gemma offer a way to explore how AI can support language learning and natural conversations. Open-source frameworks also make it possible to build and customise the surrounding application without starting everything from scratch.

I am using a hosted inference approach to make development practical without needing to run a large model on my own hardware. This is an important distinction: using an open-weight model through a hosted API is different from running the model locally.

For me, this project is about learning how to turn AI capabilities into something useful for a real problem—and making the result accessible through a simple web application.

The Lazy Payoff

Fluently aims to reduce the friction between wanting to improve English and actually practising it.

By bringing learning activities together in one place and offering AI assistance, the project is designed to make it easier to maintain a routine without spending extra effort organising every session.

The intended payoff is simple: less friction, more consistent practice, and more time spent improving communication skills.

Prize Categories

Gemma: If the deployed application uses a Gemma model as described above.

devchallenge #hf26challenge

Hacktoberfest Weekend Challenge: Build for a Friend

palakshi verma — Mon, 05 Oct 2026 06:22:46 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built DevDraft, an AI-powered communication agent designed for a friend who recently started their first software engineering internship.

Like many junior developers, my friend suffers from intense imposter syndrome when communicating in professional engineering channels. They were spending 30–45 minutes drafting a single daily standup update or a GitHub Pull Request description. The engineering work wasn't the bottleneck—it was the anxiety of getting the tone, structure, and technical jargon right in front of senior engineers.

DevDraft acts as a patient "Senior Developer" partner. My friend can dump their raw, chaotic, unfiltered notes into the app (e.g., "i fixed the login bug but the css is still weird so I didn't push that part"). The agent instantly transforms this into a crisp, confident, and professional Standup update or a perfectly structured PR template, stripping away apologetic language and formatting it for Slack or GitHub.

Demo

Live Demo: https://devdraft.onrender.com/

(Note: The live demo uses the Groq API for speed, but the app is built to be run locally!)

Code

https://github.com/PalakshiVerma/DevDraft

How I Built It

I decoupled the architecture into a Streamlit frontend and a FastAPI backend to keep the LLM logic flexible.

At the core of the project is the Meta Llama 3.1 (8B) open-weight model. The FastAPI backend receives the raw text and injects it into highly specific system prompts designed to mimic a confident senior engineer.

I deployed the FastAPI backend as a Web Service on Render. Render's automated deployment from GitHub made it incredibly easy to ship the API in minutes, providing a stable endpoint for the Streamlit UI to interact with.

Why Does Open Innovation Matter?

For an intern, using closed AI APIs is a massive security risk. Pasting proprietary company code or internal architecture notes into a closed cloud AI violates almost every company NDA.

This is where open innovation completely outshines closed systems. I built DevDraft with a "Strictly Local" mode. Because it utilizes open-weight models, my friend can run the exact same application entirely offline using Ollama.

By running Llama 3.1 locally, my friend gets senior-level code review and communication polish without a single line of their company's proprietary code ever leaving their laptop. A closed API would make this tool unusable for its intended audience; open-source AI makes it secure, free, and completely private.

Prize Categories

Best Use of Render: The core AI inference API (built with FastAPI) is hosted and deployed using Render's Web Services, acting as the bridge between the UI and the open-weight LLM.

DEV Community: palakshi verma

Fluently — Your Personal English-Learning Companion

What I Built

Demo

Code

How I Built It

Why Does Open Innovation Matter?

The Lazy Payoff

Prize Categories

devchallenge #hf26challenge

Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

Demo

Code

How I Built It

Why Does Open Innovation Matter?

Prize Categories