You do not need a new project — you need a new angle
① FIND an idea → ② MAKE it new → ③ BUILD it free → ④ WRITE it → ⑤ DEFEND it
Your professor has seen the same five projects. This reply shows where ideas live, how to twist an old one into something new, and how to build, write and defend it for free. No money, no lab, no permission needed.
Where to find ideas
| Place | What you get |
|---|---|
| build-your-own-x | build guides: database, OS, renderer |
| project-based-learning | full builds, sorted by language |
| Devpost · Devfolio · lablab.ai | hackathon projects, winners marked |
| Hackaday.io | open-hardware builds |
| awesome-public-datasets · awesome | dataset and list indexes |
| GitHub topic student-project | what classmates already shipped |
| University project galleries | real final-year titles |
github.com/codecrafters-io/build-your-own-x
github.com/practical-tutorials/project-based-learning
devpost.com/software
lablab.ai/apps
hackaday.io/projects
github.com/awesomedata/awesome-public-datasets
How to make an old idea new
- Swap the data — same method, new corpus
- Move a method across fields — NLP trick onto sensor data
- Add a constraint — run it on a microcontroller
- Audit a model — bias or robustness study
- Reproduce a result — and report where it breaks
- Make the dataset — the benchmark is the paper
- Add a human — labelling becomes the research
- Go real-time — streaming, on-device, offline
- Open the world — unseen classes, unseen domains
Data almost nobody uses
| Source | Project it enables |
|---|---|
| NOAA space weather | storm predictor |
| CelesTrak orbits | satellite pass tracker |
| NASA FIRMS fires | wildfire alerts |
| Copernicus Sentinel | land change detector |
| Global Fishing Watch | dark-vessel detector |
| ICIJ Offshore Leaks | ownership graph |
| USGS + Raspberry Shake | earthquake detector |
| IODA + Cloudflare Radar + OONI | shutdown map |
| Lichess database | engine training |
| Cricsheet + StatsBomb + FastF1 | sports analytics |
| Zooniverse | citizen-science labels |
services.swpc.noaa.gov/json
celestrak.org/NORAD/elements
firms.modaps.eosdis.nasa.gov/api
dataspace.copernicus.eu
globalfishingwatch.org/our-apis
offshoreleaks.icij.org/pages/database
earthquake.usgs.gov/earthquakes/feed/v1.0/geojson.php
ioda.inetintel.cc.gatech.edu
database.lichess.org
cricsheet.org/downloads
dataportals.org
Pick one in ten minutes
- one way from the nine
- one data source that looks interesting
- two papers, read their future-work lines
- write your one-sentence claim
- start building
🛠 Build it free
- GPU: Kaggle 30 hours a week · Colab T4 · Lightning 80 hours · Modal 30 dollars a month
- Demo host: Cloudflare Pages · Streamlit Cloud · Gradio share
- No hardware: Wokwi · Tinkercad Circuits · EDA Playground · Gazebo · Sim2Bot
- CAD and sim: KiCad · FreeCAD · OpenSCAD · Onshape free · Octave
- Scaffold: cookiecutter-data-science · cookiecutter-modern-ml · Paper2Code
- Polish: free .me domain · uptime page · one-click demo link
📄 Write and publish
- Papers free: Unpaywall · OpenAlex · Semantic Scholar · CORE · DOAJ
- Track the field: Connected Papers · ResearchRabbit · Litmaps
- Templates: Overleaf IEEE · ACM acmart · Springer lncs · Elsevier elsarticle
- Write: Zotero · LanguageTool · Inkscape · draw.io
- Publish free: arXiv or Zenodo DOI · student journals · IEEE Student Paper Contest · ACM SRC
- Check a venue: Think.Check.Submit
- Stay clean: cite your own earlier work · add an AI-use line
The forum already covered the publishing path: onehack.st/t/need-advice-fast-track-publishing-in-ieee-or-reputed-research-journals/323307
Defend it
“I did not invent a new architecture — I moved a known method onto data from [source], where it had never been tested, and I report both where it holds and where it breaks.”
Six words that make you sound like a researcher
- novelty — a new result, setting or artefact
- baseline — what you must beat
- ablation — remove a piece, show its value
- SOTA — the best published number
- preprint — a public draft with a DOI
- reproducibility — someone can rerun you
!