OpenFunds

A Streamlit crowdfunding app where creators post campaigns with a Bitcoin testnet address. A small, untested model estimates a campaign's chance of success, and sentence embeddings flag near-duplicates. A four-person course project, later the subject of a university-guided internship report.

When: Apr – Oct 2025
Status: Built
Team: RV University Agile course project: a team of four with a faculty guide.
Role: One of four listed co-authors.

Progress55%

A working Streamlit app. The prediction model has no evaluation, and the Bitcoin part only checks the format of an address.

What comes next

Evaluate the success predictor on real campaign data, or remove it.
Build the payment path, or describe the app as address validation only.
Capture clean screenshots of the running app.

Not measured

accuracy of the campaign success predictor

There is no evaluation code. The model uses three features, and the database holds 10 campaigns.

Audit of the project code (models/ml_predictor.py, data/campaigns.db)

Format check only

what the app does with a Bitcoin testnet address

No payment moves and nothing is decentralized. The wallet connection was an interface mock-up.

Audit of the project code (pages/2_Create_Campaign.py) and the sprint logbook

What it is

OpenFunds is a crowdfunding web app. Creators post campaigns with a Bitcoin testnet address, backers browse and back them, and two pieces of machine learning sit on top: a model that estimates a campaign's chance of success, and sentence embeddings that flag campaigns that look like copies of one another. It began in April 2025 as a four-person project for an Agile software engineering and DevOps course at RV University.

What was built

A multi-page Streamlit app with user accounts and SQLite storage, for browsing, creating and backing campaigns. A success predictor that combines logistic regression and a random forest. Near-duplicate detection with MiniLM sentence embeddings. Validation and display of Bitcoin testnet addresses. A Dockerfile and Kubernetes deployment and service files. And the course paperwork: a sprint logbook, an Agile report and a DevOps report.

What it does not claim

The code is lighter than the reports written about it. The predictor uses three features, has no evaluation code, and the database holds 10 campaigns, so this page quotes no accuracy. The Bitcoin part checks that an address is well formed and nothing more: no payment moves, and nothing here is decentralized. A wallet connection existed only as an interface mock-up.

The internship

In July 2025 the project became the subject of my B.Tech honors internship report. That internship was guided by the university. It was not a placement at a company, and it is not listed as work experience on this site.

What I'd change

Evaluate the predictor on real campaign data before it shows a probability to anyone, or take it out. And either build the payment path or describe the app as what it is: address validation.

One question

The answer is on this page. Get it right and it goes in your passport on the map.

Challenge

OpenFunds has a model that predicts whether a campaign will succeed. How accurate is it?

Show the answer

Nobody measured it. It was never evaluated. There is no evaluation code, the model uses three features, and the database holds 10 campaigns. Source: the project audit of November 2026.

Stack: Python
Streamlit
SQLite
scikit-learn
sentence-transformers (MiniLM)
Docker
Kubernetes manifests