twangodev/devpost-hacks

---
pretty_name: Devpost Hackathon Projects
license: other
language:
- en
size_categories:
- 1K<n<10K
task_categories:
- text-classification
- text-generation
- summarization
tags:
- hackathon
- devpost
- github
- projects
configs:
- config_name: all
default: true
data_files:
- split: train
path: data/all/train.parquet
- config_name: cal-hacks-12-0
data_files:
- split: train
path: data/cal-hacks-12-0/train.parquet
- config_name: hackgt-12
data_files:
- split: train
path: data/hackgt-12/train.parquet
- config_name: hacktech-by-caltech-2026
data_files:
- split: train
path: data/hacktech-by-caltech-2026/train.parquet
- config_name: madhacks
data_files:
- split: train
path: data/madhacks/train.parquet
- config_name: madhacks-fall-2025
data_files:
- split: train
path: data/madhacks-fall-2025/train.parquet
- config_name: pennapps-xxv
data_files:
- split: train
path: data/pennapps-xxv/train.parquet
- config_name: treehacks-2024
data_files:
- split: train
path: data/treehacks-2024/train.parquet
- config_name: treehacks-2025
data_files:
- split: train
path: data/treehacks-2025/train.parquet
- config_name: treehacks-2026
data_files:
- split: train
path: data/treehacks-2026/train.parquet
---

# devpost-hacks

A collection of hackathon project submissions scraped from [Devpost](https://devpost.com),
enriched with the README files of any GitHub repos linked from each project.

Intended for **research use only** (e.g. evaluating LLM judges, training project-summarization
models, studying what wins hackathons.)

## Configurations

| Config | Rows | Winners | With READMEs |
|---|---:|---:|---:|
| `all` (default) | 2222 | 363 | 1358 |
| `cal-hacks-12-0` | 694 | 100 | 352 |
| `hackgt-12` | 272 | 25 | 171 |
| `hacktech-by-caltech-2026` | 61 | 0 | 46 |
| `madhacks` | 55 | 10 | 32 |
| `madhacks-fall-2025` | 111 | 13 | 77 |
| `pennapps-xxv` | 98 | 25 | 63 |
| `treehacks-2024` | 319 | 58 | 192 |
| `treehacks-2025` | 248 | 70 | 185 |
| `treehacks-2026` | 364 | 62 | 240 |

`hacktech-by-caltech-2026` has zero winners because results were not yet announced
on Devpost at scrape time.

## Schema

| Field | Type | Notes |
|---|---|---|
| `project_id` | string | First 12 hex chars of `sha1(url)` |
| `hackathon` | string | Hackathon slug (matches the config name) |
| `url` | string | Devpost project URL |
| `title` | string | |
| `tagline` | string | One-line summary on Devpost |
| `description` | string | Full Devpost project write-up |
| `built_with` | list&lt;string&gt; | Tags the team used (`react`, `python`, ...) |
| `video_link` | string | Demo video URL if any |
| `other_links` | list&lt;string&gt; | All non-video links from the project page |
| `results` | string | Award label (`Winner`, `Did Not Place`, etc.) or `null` |
| `is_winner` | bool | True iff Devpost flagged any award as a "winner" |
| `readmes` | list&lt;struct&gt; | GitHub READMEs for repos linked under `other_links` |

The `readmes` struct fields:

- `repo` — `owner/repo`
- `content` — README markdown, truncated to 6000 chars (with `[... truncated]` marker)
- `truncated` — bool

## Loading

```python
from datasets import load_dataset

ds = load_dataset("twangodev/devpost-hacks") # all hackathons
ds = load_dataset("twangodev/devpost-hacks", "treehacks-2026") # one hackathon
```

## Sources & licensing

This dataset aggregates content from two upstream sources, neither of which has a
single uniform license:

1. Devpost project pages - Project text (title, tagline, description, links, results)
is authored by the participating teams. We include this content here under
a fair-use / research-use rationale.
2. README files: Each README inherits the license of its repository.

Because of (2) in particular, we cannot apply a uniform open license to this
dataset.

## Takedown & Removal Requests

If you are the author of a project (or a maintainer of a linked repo) and would
like your content removed from this dataset, email **[e-mail]** with
the project URL or repo slug. We will remove it from the next release.