Devpost
Participate in our public hackathons
Devpost for Teams
Access your company's private hackathons
Grow your developer ecosystem and promote your platform
Drive innovation, collaboration, and retention within your organization
By use case
Blog
Insights into hackathon planning and participation
Customer stories
Inspiration from peers and other industry leaders
Planning guides
Best practices for planning online and in-person hackathons
Webinars & events
Upcoming events and on-demand recordings
Help desk
Common questions and support documentation
Sort by:
A bio-inspired dual-process cognitive controller that gives open-weight AI human-like deliberate reasoning, zero catastrophic forgetting, and 50% compute savings on consumer hardware.
Argus turns one photo or a full shoot into inspectable 3D and PBR material prototypes. Evidence checks show what works, what needs fixing, and why—powered by NVIDIA AI in Nebius.
An autonomous AI research engineer that turns a research question into experiments, trains/finetunes models on GPUs, evaluates them, discovers failures, and produces reproducible research reports.
I trained ODIN Pocket from scratch on an RTX 4070. It’s a 49.3M-parameter language model you can run on a CPU to generate text and inspect the probability of each token.
A 49.5M-parameter language model trained from scratch for $14 of GPU time that beats GPT-2-124M on ARC-Easy. Tells you plainly which of its other results are just noise.
HelixDepth tests whether reusing six Transformer blocks across twelve stages can improve next-token prediction in a 41-million-parameter language model.
Convert videos into photorealistic 3D Gaussian & Triangle Splats with an embodied Spatial AI Guide powered by NVIDIA Nemotron on Nebius Token Factory and Tavily web grounding in WebXR.
Small by design. Trained from scratch. Measured by evidence.
An AI-powered space signal processing pipeline using NVIDIA CUDA to filter raw telemetry noise and isolate cosmic vectors.
This is a subsea ROV station built for low visibility and autonomous search and recovery. This unit has been designed with plug and play option , no downloads or internet connection needed.
Three seconds of speech can clone a voice. ProofVoice detects AI-generated speech in files and live streams, shows the evidence behind every verdict, and says when it isn't sure.
a radio silent drone that uses onboard AI to find people, confirm them with a vision-language model, and pinpoint where they are, using only a camera and IMU.
Have you ever thought "Man, how cool would it be if fly could play drums??" Well now... this one CAN!!!!! THIS FLY PLAYS DRUMS!!!! You didn't think it was possible, and yet, here it is.
An eye for the cane that guides indoor and outdoor navigation for poorly sighted people, along with giving them 360 degree vision and having 24/7 agentic assistance whenever and wherever.
A sovereign, local AI ecosystem that breaks away from the cloud to bring private, fast, and autonomous intelligence straight to the hardware.
Breaking communication barriers with real-time, bidirectional Taiwan Sign Language translation powered by Gemma AI and 3D animation.
A from-scratch C++17/CUDA machine learning runtime with tensors, autograd, CPU/GPU backends, mixed precision, neural-network layers, checkpointing, and distributed training.
VAnta — Smarter GPU Scheduling, Faster Results.
Subtítulos y traducción en vivo para todas las salas de una conferencia, en una GPU de escritorio: 32 salas medidas, sin operadores, sin pagar por minuto, open source.
Open-source real-time captions for conferences with 30+ parallel sessions — local or cloud ASR, fast local NMT translation, streamed live to phones, OBS, and venue screens.
MORICE turns an AI model from something you talk to into something that can actually work with you.
Every voice. Every language. Everyone included. EventLyra turns a live talk into open-source captions the whole room can follow, in real time.
A 43M-parameter language model built from scratch on one 6 GB laptop GPU: our own tokenizer, data pipeline and trainer, 1B tokens in 15.8 hours, with every reported number verified.
Open-source bilingual live captions that scale by stage, not by audience.
