Discover the innovative AI applications that have been built during lablab.ai hackathons. Explore our AI Hackathon projects and get inspired to create your own!
Garuda Linux is an Arch-based Linux distribution known for its striking visual design, performance-focused tweaks (like BTRFS with automatic snapshots and Zen kernel), and a strong focus on gaming.
Uses Gemma 4 to classify complex vs. simple requests, and routes them to a local LLM / cloud provider as needed.
Quantivo AI (BOA) generates AI-powered Business Opportunity Analysis reports by combining local market data, sentiment analysis, and SWOT insights to help entrepreneurs validate and grow their business ideas.
SovereignSwarm is an autonomous video captioning engine using Fireworks AI and OpenCV to extract chronological frames and generate structured, multi-tone captions (Formal, Sarcastic, Humorous Tech, Everyday) in an enterprise-ready JSON format.
A project showing that neither pure LLMs or pure ML is the best way to approach autonomous trading.
Aura is a sovereign, neuro-symbolic AI operating system that turns natural-language intent into transparent, governed action through local parsing, agent collaboration, bounded experimentation, verifiable learning, and human-controlled decision gates.
AI LifeOS is an AI-powered platform that helps students learn, build skills, explore career paths, and prepare for jobs through personalized roadmaps, AI tutoring, and progress tracking.
NarraFlow is a multi-agent AI storytelling platform that transforms simple prompts into immersive stories. Specialized AI agents collaborate to plan plots, develop characters, maintain consistency, and generate engaging narratives in real time
TRANSDUCER watches any video clip and captions it in four voices - formal, sarcastic, humorous_tech, humorous_non_tech. One 1024px look, four persona briefs, grounded in what's on screen
This project is a simple video captioner. Uses a VLM to extract the text from a video and another model to transform the description in the required styles. As a fallback, it contains a fine-tunned local model based on Gemma 4.
AIVE (Artificial Intelligence Venture Engine) is a Cognitive Discovery Operating System that transforms documents, research papers, and data into structured knowledge, evidence-backed insights, dynamic reasoning, and actionable innovation opportunities.
ResearchOS is a multi-agent AI research laboratory that autonomously retrieves academic evidence, analyzes literature, identifies research gaps, and generates evidence-grounded research reports.
SOAP Copilot turns raw doctor-patient conversations into structured SOAP notes, ICD-10 codes, and patient-friendly summaries in seconds, using a 3-agent Llama 3.3 70B pipeline built and fine-tuned on AMD hardware.
A reinforcement-learning router that learns which AI model to call for each query, cutting cost by ~93% without ever hardcoding a routing rule.
Thymus is a lightweight hybrid token-efficient router designed to maximize accuracy while minimizing token costs in multi‑task LLM pipelines. It dynamically routes user queries across local and remote models on LLM providers.
An agentic pipeline that takes any CUDA repo and autonomously ports it to ROCm/HIP — analyze → transpile → compile → test → benchmark on real AMD GPUs — with an LLM repair loop fixing everything the mechanical tools can't.
A privacy-first classroom AI platform that routes sensitive work to local edge systems and eligible anonymized analysis to Fireworks AI, helping teachers make faster, safer instructional decisions even with unreliable internet.
Taskly classifies incoming tasks into 8 categories (QA, math, code, NLP) and routes each to the optimal Fireworks AI model with a tuned prompt — maximizing accuracy while minimizing token usage in a fully Dockerized pipeline.
A video captioning agent that watches a clip once, then rewrites what it saw as Formal, Sarcastic, or Humorous captions
CareerLaunch AI: An intelligent workspace architect that synthesizes personalized technical roadmaps, project directories, and AMD-optimized hardware configurations using Llama 3 AI via Fireworks.
A Natively AI orchestrated , Adaptive Routing Agent integrated with Fireworks AI . It profiles LLM tasks and routes them to the optimal AI model using task profiling and utility scoring to reduce cost and latency while maximizing output quality.
Every interaction makes your AI team smarter. Agents Terminal V3 is a self-improving orchestration platform. WhatsApp messages flow through a 7-intent classifier, get captured and automatically upgraded every Saturday — smarter week after week.
3DP Agent analyzes STL files before printing, catching wall-thickness failures and unsupported overhangs invisible to the eye. Five specialized AI agents vote on a consensus printability score, with inference running on AMD ROCm via vLLM.
Aprendia turns any topic into a playable, gamified course built on learning science: 13 exercise types, FSRS spaced repetition, exam readiness and a cited RAG tutor all AI on AMD: Fireworks open models + self-hosted Gemma 3 27B on an MI300X.
Claption turns short videos into accurate captions in four tones: formal, sarcastic, humorous-tech, and humorous-non-tech, using Fireworks AI with a judge-and-repair loop.
Vision-based video captioning pipeline using Gemma 4 31B. Analyzes video frames,nbuilds scene understanding, generates 4 caption styles with automatic quality scoring.
A context-aware Windows window manager: say what you're doing in plain English, and an AMD-hosted LLM instantly rearranges your open windows around your intent.
Vektor is a deterministic Guardian Runtime that intercepts autonomous AI tool calls before execution. Because it evaluates actions against enterprise policies without language models, it prevents unsafe behaviour and logs verifiable proof.
Macro Conflict Radar answers macro questions the way an auditor would — every number produced by deterministic code, every sentence traced back to a stored fact. The whole point of radar is to catch the conflicts where they usually agree.
TipFork is an AI menu agent: snap a restaurant menu and it extracts every dish and price, translates dish names, generates AI visuals of each dish, and estimates tax and tip with a fair per-person split, reconciled against the real receipt.